1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

434

435

436

437

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

521

522

523

524

525

526

527

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

552

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

580

581

582

583

584

585

586

587

588

589

590

591

592

593

594

595

596

597

598

599

600

601

602

603

604

605

606

607

608

609

610

611

612

613

614

615

616

617

618

619

620

621

622

623

624

625

626

627

628

629

630

631

632

633

634

635

636

637

638

639

640

641

642

643

644

645

646

647

648

649

650

651

652

653

654

655

656

657

658

659

660

661

662

663

664

665

666

667

668

669

670

671

672

673

674

675

676

677

678

679

680

681

682

683

684

685

686

687

688

689

690

691

692

693

694

695

696

697

698

699

700

701

702

703

704

705

706

707

708

709

710

711

712

713

714

715

716

717

718

719

720

721

722

723

724

725

726

727

728

729

730

731

732

733

734

735

736

737

738

739

740

741

742

743

744

745

746

747

748

749

750

751

752

753

754

755

756

757

758

759

760

761

762

763

764

765

766

767

768

769

770

771

772

773

774

775

776

777

778

779

780

781

782

783

784

785

786

787

788

789

790

791

792

793

794

795

796

797

798

799

800

801

802

803

804

805

806

807

808

809

810

811

812

813

814

815

816

817

818

819

820

821

822

823

824

825

826

827

828

829

830

831

832

833

834

835

836

837

838

839

840

841

842

843

844

845

846

847

848

849

850

851

852

853

854

855

856

857

858

859

860

861

862

863

864

865

866

867

868

869

870

871

872

873

874

875

876

877

878

879

880

881

882

883

884

885

886

887

888

889

890

891

892

893

894

895

896

897

898

899

900

901

902

903

904

905

906

907

908

909

910

911

912

913

914

915

916

917

918

919

920

921

922

923

924

925

926

927

928

929

930

931

932

933

934

935

936

937

938

939

940

941

942

943

944

945

946

947

948

949

950

951

952

953

954

955

956

957

958

959

960

961

962

963

964

965

966

967

968

969

970

971

972

973

974

975

976

977

978

979

980

981

982

983

984

985

986

987

988

989

990

991

992

993

994

995

996

997

998

999

1000

1001

1002

1003

1004

1005

1006

1007

1008

1009

1010

1011

1012

1013

1014

1015

1016

1017

1018

1019

1020

1021

1022

1023

1024

1025

1026

1027

1028

1029

1030

1031

1032

1033

1034

1035

1036

1037

1038

1039

1040

1041

1042

1043

1044

1045

1046

1047

1048

1049

1050

1051

1052

1053

1054

1055

1056

1057

1058

1059

1060

1061

1062

1063

1064

1065

1066

1067

1068

1069

1070

1071

1072

1073

1074

1075

1076

1077

1078

1079

1080

1081

1082

1083

1084

1085

1086

1087

1088

1089

1090

1091

1092

1093

1094

1095

1096

1097

1098

1099

1100

1101

1102

1103

1104

1105

1106

1107

1108

1109

1110

1111

1112

1113

1114

1115

1116

1117

1118

1119

1120

1121

1122

1123

1124

1125

1126

1127

1128

1129

1130

1131

1132

1133

1134

1135

1136

1137

1138

1139

1140

1141

1142

1143

1144

1145

1146

1147

1148

1149

1150

1151

1152

1153

1154

1155

1156

1157

1158

1159

1160

1161

1162

1163

1164

1165

1166

1167

1168

1169

1170

1171

1172

1173

1174

1175

1176

1177

1178

1179

1180

1181

1182

1183

1184

1185

1186

1187

1188

1189

1190

1191

1192

1193

1194

1195

1196

1197

1198

1199

1200

1201

1202

1203

1204

1205

1206

1207

1208

1209

1210

1211

1212

1213

1214

1215

1216

1217

1218

1219

1220

1221

1222

1223

1224

1225

1226

1227

1228

1229

1230

1231

1232

1233

1234

1235

1236

1237

1238

1239

1240

1241

1242

1243

1244

1245

1246

1247

1248

1249

1250

1251

1252

1253

1254

1255

1256

1257

1258

1259

1260

1261

1262

1263

1264

1265

1266

1267

1268

1269

1270

1271

1272

1273

1274

1275

1276

1277

1278

1279

1280

1281

1282

1283

1284

1285

1286

1287

1288

1289

1290

1291

1292

1293

1294

1295

1296

1297

1298

1299

1300

1301

1302

1303

1304

1305

1306

1307

1308

1309

1310

1311

1312

1313

1314

1315

1316

1317

1318

1319

1320

1321

1322

1323

1324

1325

1326

1327

1328

1329

1330

1331

1332

1333

1334

1335

1336

1337

1338

1339

1340

1341

1342

1343

1344

1345

1346

1347

1348

1349

1350

1351

1352

1353

1354

1355

1356

1357

1358

1359

1360

1361

1362

1363

1364

1365

1366

1367

1368

1369

1370

1371

1372

1373

1374

1375

1376

1377

1378

1379

1380

1381

1382

1383

1384

1385

1386

1387

1388

1389

1390

1391

1392

1393

1394

1395

1396

1397

1398

1399

1400

1401

1402

1403

1404

1405

1406

1407

1408

1409

1410

1411

1412

1413

1414

1415

1416

1417

1418

1419

1420

1421

1422

1423

1424

1425

1426

1427

1428

1429

1430

1431

1432

1433

1434

1435

1436

1437

1438

1439

1440

1441

1442

1443

1444

1445

1446

1447

1448

1449

1450

1451

1452

1453

1454

1455

1456

1457

1458

1459

1460

1461

1462

1463

1464

1465

1466

1467

1468

1469

1470

1471

1472

1473

1474

1475

1476

1477

1478

1479

1480

1481

1482

1483

1484

1485

1486

1487

1488

1489

1490

1491

1492

1493

1494

1495

1496

1497

1498

1499

1500

1501

1502

1503

1504

1505

1506

1507

1508

1509

1510

1511

1512

1513

1514

1515

1516

1517

1518

1519

1520

1521

1522

1523

1524

1525

1526

1527

1528

1529

1530

1531

1532

1533

1534

1535

1536

1537

1538

1539

1540

1541

1542

1543

1544

1545

1546

1547

1548

1549

1550

1551

1552

1553

1554

1555

1556

1557

1558

1559

1560

1561

1562

1563

1564

1565

1566

1567

1568

1569

1570

1571

1572

1573

1574

1575

1576

1577

1578

1579

1580

1581

1582

1583

1584

1585

1586

1587

1588

1589

1590

1591

1592

1593

1594

1595

1596

1597

1598

1599

1600

1601

1602

1603

1604

1605

1606

1607

1608

1609

1610

1611

1612

1613

1614

1615

1616

1617

1618

1619

1620

1621

1622

1623

1624

1625

1626

1627

1628

1629

1630

1631

1632

1633

1634

1635

1636

1637

1638

1639

1640

1641

1642

1643

1644

1645

1646

1647

1648

1649

1650

1651

1652

1653

1654

1655

1656

1657

1658

1659

1660

1661

1662

1663

1664

1665

1666

1667

1668

1669

1670

1671

1672

1673

1674

1675

1676

1677

1678

1679

1680

1681

1682

1683

1684

1685

1686

1687

1688

1689

1690

1691

1692

1693

1694

1695

1696

1697

1698

1699

1700

1701

1702

1703

1704

1705

1706

1707

1708

1709

1710

1711

1712

1713

1714

1715

1716

1717

1718

1719

1720

1721

1722

1723

1724

1725

1726

1727

1728

1729

1730

1731

1732

1733

1734

1735

1736

1737

1738

1739

1740

1741

1742

1743

1744

1745

1746

1747

1748

1749

1750

1751

1752

1753

1754

1755

1756

1757

1758

1759

1760

1761

1762

1763

1764

1765

1766

1767

1768

1769

1770

1771

1772

1773

1774

1775

1776

1777

1778

1779

1780

1781

1782

1783

1784

1785

1786

1787

1788

1789

1790

1791

1792

1793

1794

1795

1796

1797

1798

1799

1800

1801

1802

1803

1804

1805

1806

1807

1808

1809

1810

1811

1812

1813

1814

1815

1816

1817

1818

1819

1820

1821

1822

1823

1824

1825

1826

1827

1828

1829

1830

1831

1832

1833

1834

1835

1836

1837

1838

1839

1840

1841

1842

1843

1844

1845

1846

1847

1848

1849

1850

1851

1852

1853

1854

1855

1856

1857

1858

1859

1860

1861

1862

1863

1864

1865

1866

1867

1868

1869

1870

1871

1872

1873

1874

1875

1876

1877

1878

1879

1880

1881

1882

1883

1884

1885

1886

1887

1888

1889

1890

1891

1892

1893

1894

1895

1896

1897

1898

1899

1900

1901

1902

1903

1904

1905

1906

1907

1908

1909

1910

1911

1912

1913

1914

1915

1916

1917

1918

1919

1920

1921

1922

1923

1924

1925

1926

1927

1928

1929

1930

1931

1932

1933

1934

1935

1936

1937

1938

1939

1940

1941

1942

1943

1944

1945

1946

1947

1948

1949

1950

1951

1952

1953

1954

1955

1956

1957

1958

1959

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021

2022

2023

2024

2025

2026

2027

2028

2029

2030

2031

2032

2033

2034

2035

2036

2037

2038

2039

2040

2041

2042

2043

2044

2045

2046

2047

2048

2049

2050

2051

2052

2053

2054

2055

2056

2057

2058

2059

2060

2061

2062

2063

2064

2065

2066

2067

2068

2069

2070

2071

2072

2073

2074

2075

2076

2077

2078

2079

2080

2081

2082

2083

2084

2085

2086

2087

2088

2089

2090

2091

2092

2093

2094

2095

2096

2097

2098

2099

2100

2101

2102

2103

2104

2105

2106

2107

2108

2109

2110

2111

2112

2113

2114

2115

2116

2117

2118

2119

2120

2121

2122

2123

2124

2125

2126

2127

2128

2129

2130

2131

2132

2133

2134

2135

2136

2137

2138

2139

2140

2141

2142

2143

2144

2145

2146

2147

2148

2149

2150

2151

2152

2153

2154

2155

2156

2157

2158

2159

2160

2161

2162

2163

2164

2165

2166

2167

2168

2169

2170

2171

2172

2173

2174

2175

2176

2177

2178

2179

2180

2181

2182

2183

2184

2185

2186

2187

2188

2189

2190

2191

2192

2193

2194

2195

2196

2197

2198

2199

2200

2201

2202

2203

2204

2205

2206

2207

2208

2209

2210

2211

2212

2213

2214

2215

2216

2217

2218

2219

2220

2221

2222

2223

2224

2225

2226

2227

2228

2229

2230

2231

2232

2233

2234

2235

2236

2237

2238

2239

2240

2241

2242

2243

2244

2245

2246

2247

2248

2249

2250

2251

2252

2253

2254

2255

2256

2257

2258

2259

2260

2261

2262

2263

2264

2265

2266

2267

2268

2269

2270

2271

2272

2273

2274

2275

2276

2277

2278

2279

2280

2281

2282

2283

2284

2285

2286

2287

2288

2289

2290

2291

2292

2293

2294

2295

2296

2297

2298

2299

2300

2301

2302

2303

2304

2305

2306

2307

2308

2309

2310

2311

2312

2313

2314

2315

2316

2317

2318

2319

2320

2321

2322

2323

2324

2325

2326

2327

2328

2329

2330

2331

2332

2333

2334

2335

2336

2337

2338

2339

2340

2341

2342

2343

2344

2345

2346

2347

2348

2349

2350

2351

2352

2353

2354

2355

2356

2357

2358

2359

2360

2361

2362

2363

2364

2365

2366

2367

2368

2369

2370

2371

2372

2373

2374

2375

2376

2377

2378

2379

2380

2381

2382

2383

2384

2385

2386

2387

2388

2389

2390

2391

2392

2393

2394

2395

2396

2397

2398

2399

2400

2401

2402

2403

2404

2405

2406

2407

2408

2409

2410

2411

2412

2413

2414

2415

2416

2417

2418

2419

2420

2421

2422

2423

2424

2425

2426

2427

2428

2429

2430

2431

2432

2433

2434

2435

2436

2437

2438

2439

2440

2441

2442

2443

2444

2445

2446

2447

2448

2449

2450

2451

2452

2453

2454

2455

2456

2457

2458

2459

2460

2461

2462

2463

2464

2465

2466

2467

2468

2469

2470

2471

2472

2473

2474

2475

2476

2477

2478

2479

2480

2481

2482

2483

2484

2485

2486

2487

2488

2489

2490

2491

2492

2493

2494

2495

2496

2497

2498

2499

2500

2501

2502

2503

2504

2505

2506

2507

2508

2509

2510

2511

2512

2513

2514

2515

2516

2517

2518

2519

2520

2521

2522

2523

2524

2525

2526

2527

2528

2529

2530

2531

2532

2533

2534

2535

2536

2537

2538

2539

2540

2541

2542

2543

2544

2545

2546

2547

2548

2549

2550

2551

2552

2553

2554

2555

2556

2557

2558

2559

2560

2561

2562

2563

2564

2565

2566

2567

2568

2569

2570

2571

2572

2573

2574

2575

2576

2577

2578

2579

2580

2581

2582

2583

2584

2585

2586

2587

2588

2589

2590

2591

2592

2593

2594

2595

2596

2597

2598

2599

2600

2601

2602

2603

2604

2605

2606

2607

2608

2609

2610

2611

2612

2613

2614

2615

2616

2617

2618

2619

2620

2621

2622

2623

2624

2625

2626

2627

2628

2629

2630

2631

2632

2633

2634

2635

2636

2637

2638

2639

2640

2641

2642

2643

2644

2645

2646

2647

2648

2649

2650

2651

2652

2653

2654

2655

2656

2657

2658

2659

2660

2661

2662

2663

2664

2665

2666

2667

2668

2669

2670

2671

2672

2673

2674

2675

2676

2677

2678

2679

2680

2681

2682

2683

2684

2685

2686

2687

2688

2689

2690

2691

2692

2693

2694

2695

2696

2697

2698

2699

2700

2701

2702

2703

2704

2705

2706

2707

2708

2709

2710

2711

2712

2713

2714

2715

2716

2717

2718

2719

2720

2721

2722

2723

2724

2725

2726

2727

2728

2729

2730

2731

2732

2733

2734

2735

2736

2737

2738

2739

2740

2741

2742

2743

2744

2745

2746

2747

2748

2749

2750

2751

2752

2753

2754

2755

2756

2757

2758

2759

2760

2761

2762

2763

2764

2765

2766

2767

2768

2769

2770

2771

2772

2773

2774

2775

2776

2777

2778

2779

2780

2781

2782

2783

2784

2785

2786

2787

2788

2789

2790

2791

2792

2793

2794

2795

2796

2797

2798

2799

2800

2801

2802

2803

2804

2805

2806

2807

2808

2809

2810

2811

2812

2813

2814

2815

2816

2817

2818

2819

2820

2821

2822

2823

2824

2825

2826

2827

2828

2829

2830

2831

2832

2833

2834

2835

2836

2837

2838

2839

2840

2841

2842

2843

2844

2845

2846

2847

2848

2849

2850

2851

2852

2853

2854

2855

2856

2857

2858

2859

2860

2861

2862

2863

2864

2865

2866

2867

2868

2869

2870

2871

2872

2873

2874

2875

2876

2877

2878

2879

2880

2881

2882

2883

2884

2885

2886

2887

2888

2889

2890

2891

2892

2893

2894

2895

2896

2897

2898

2899

2900

2901

2902

2903

2904

2905

2906

2907

2908

2909

2910

2911

2912

2913

2914

2915

2916

2917

2918

2919

2920

2921

2922

2923

2924

2925

2926

2927

2928

2929

2930

2931

2932

2933

2934

2935

2936

2937

2938

2939

2940

2941

2942

2943

2944

2945

2946

2947

2948

2949

2950

2951

2952

2953

2954

2955

2956

2957

2958

2959

2960

2961

2962

2963

2964

2965

2966

2967

2968

2969

2970

2971

2972

2973

2974

2975

2976

2977

2978

2979

2980

2981

2982

2983

2984

2985

2986

2987

2988

2989

2990

2991

2992

2993

2994

2995

2996

2997

2998

2999

3000

3001

3002

3003

3004

3005

3006

3007

3008

3009

3010

3011

3012

3013

3014

3015

3016

3017

3018

3019

3020

3021

3022

3023

3024

3025

3026

3027

3028

3029

3030

3031

3032

3033

3034

3035

3036

3037

3038

3039

3040

3041

3042

3043

3044

3045

3046

3047

3048

3049

3050

3051

3052

3053

3054

3055

3056

3057

3058

3059

3060

3061

3062

3063

3064

3065

3066

3067

3068

3069

3070

3071

3072

3073

3074

3075

3076

3077

3078

3079

3080

3081

3082

3083

3084

3085

3086

3087

3088

3089

3090

3091

3092

3093

3094

3095

3096

3097

3098

3099

3100

3101

3102

3103

3104

3105

3106

3107

3108

3109

3110

3111

3112

3113

3114

3115

3116

3117

3118

3119

3120

3121

3122

3123

3124

3125

3126

3127

3128

3129

3130

3131

3132

3133

3134

3135

3136

3137

3138

3139

3140

3141

3142

3143

3144

3145

3146

3147

3148

3149

3150

3151

3152

3153

3154

3155

3156

3157

3158

3159

3160

3161

3162

3163

3164

3165

3166

3167

3168

3169

3170

3171

3172

3173

3174

3175

3176

3177

3178

3179

3180

3181

3182

3183

3184

3185

3186

3187

3188

3189

3190

3191

3192

3193

3194

3195

3196

3197

3198

3199

3200

3201

3202

3203

3204

3205

3206

3207

3208

3209

3210

3211

3212

3213

3214

3215

3216

3217

3218

3219

3220

3221

3222

3223

3224

3225

3226

3227

3228

3229

3230

3231

3232

3233

3234

3235

3236

3237

3238

3239

3240

3241

3242

3243

3244

3245

3246

3247

3248

3249

3250

3251

3252

3253

3254

3255

3256

3257

3258

3259

3260

3261

3262

3263

3264

3265

3266

3267

3268

3269

3270

3271

3272

3273

3274

3275

3276

3277

3278

3279

3280

3281

3282

3283

3284

3285

3286

3287

3288

3289

3290

3291

3292

3293

3294

3295

3296

3297

3298

3299

3300

3301

3302

3303

3304

3305

3306

3307

3308

3309

3310

3311

3312

3313

3314

3315

3316

3317

3318

3319

3320

3321

3322

3323

3324

3325

3326

3327

3328

3329

3330

3331

3332

3333

3334

3335

3336

3337

3338

3339

3340

3341

3342

3343

3344

3345

3346

3347

3348

3349

3350

3351

3352

3353

3354

3355

3356

3357

3358

3359

3360

3361

3362

3363

3364

3365

3366

3367

3368

3369

3370

3371

3372

3373

3374

3375

3376

3377

3378

3379

3380

3381

3382

3383

3384

3385

3386

3387

3388

3389

3390

3391

3392

3393

3394

3395

3396

3397

3398

3399

3400

3401

3402

3403

3404

3405

3406

3407

3408

3409

3410

3411

3412

3413

3414

3415

3416

3417

3418

3419

3420

3421

3422

3423

3424

3425

3426

3427

3428

3429

3430

3431

3432

3433

3434

3435

3436

3437

3438

3439

3440

3441

3442

3443

3444

3445

3446

3447

3448

3449

3450

3451

3452

3453

3454

3455

3456

3457

3458

3459

3460

3461

3462

3463

3464

3465

3466

3467

3468

3469

3470

3471

3472

3473

3474

3475

3476

3477

3478

3479

3480

3481

3482

3483

3484

3485

3486

3487

3488

3489

3490

3491

3492

3493

3494

3495

3496

3497

3498

3499

3500

3501

3502

3503

3504

3505

3506

3507

3508

3509

3510

3511

3512

3513

3514

3515

3516

3517

3518

3519

3520

3521

3522

3523

3524

3525

3526

3527

3528

3529

3530

3531

3532

3533

3534

3535

3536

3537

3538

3539

3540

3541

3542

3543

3544

3545

3546

3547

3548

3549

3550

3551

3552

3553

3554

3555

3556

3557

3558

3559

3560

3561

3562

3563

3564

3565

3566

3567

3568

3569

3570

3571

3572

3573

3574

3575

3576

3577

3578

3579

3580

3581

3582

3583

3584

3585

3586

3587

3588

3589

3590

3591

3592

3593

3594

3595

3596

3597

3598

3599

3600

3601

3602

3603

3604

3605

3606

3607

3608

3609

3610

3611

3612

3613

3614

3615

3616

3617

3618

3619

3620

3621

3622

3623

3624

3625

3626

3627

3628

3629

3630

3631

3632

3633

3634

3635

3636

3637

3638

3639

3640

3641

3642

3643

3644

3645

3646

3647

3648

3649

3650

3651

3652

3653

3654

3655

3656

3657

3658

3659

3660

3661

3662

3663

3664

3665

3666

3667

3668

3669

3670

3671

3672

3673

3674

3675

3676

3677

3678

3679

3680

3681

3682

3683

3684

3685

3686

3687

3688

3689

3690

3691

3692

3693

3694

3695

3696

3697

3698

3699

3700

3701

3702

3703

3704

3705

3706

3707

3708

3709

3710

3711

3712

3713

3714

3715

3716

3717

3718

3719

3720

3721

3722

3723

3724

3725

3726

3727

3728

3729

3730

3731

3732

3733

3734

3735

3736

3737

3738

3739

3740

3741

3742

3743

3744

3745

3746

3747

3748

3749

3750

3751

3752

3753

3754

3755

3756

3757

3758

3759

3760

3761

3762

3763

3764

3765

3766

3767

3768

3769

3770

3771

3772

3773

3774

3775

3776

3777

3778

3779

3780

3781

3782

3783

3784

3785

3786

3787

3788

3789

3790

3791

3792

3793

3794

3795

3796

3797

3798

3799

3800

3801

3802

3803

3804

3805

3806

3807

3808

3809

3810

3811

3812

3813

3814

3815

3816

3817

3818

3819

3820

3821

3822

3823

3824

3825

3826

3827

3828

3829

3830

3831

3832

3833

3834

3835

3836

3837

3838

3839

3840

3841

3842

3843

3844

3845

3846

3847

3848

3849

3850

3851

3852

3853

3854

3855

3856

3857

3858

3859

3860

3861

3862

3863

3864

3865

3866

3867

3868

3869

3870

3871

3872

3873

3874

3875

3876

3877

3878

3879

3880

3881

3882

3883

3884

3885

3886

3887

3888

3889

3890

3891

3892

3893

3894

3895

3896

3897

3898

3899

3900

3901

3902

3903

3904

3905

3906

3907

3908

3909

3910

3911

3912

3913

3914

3915

3916

3917

3918

3919

3920

3921

3922

3923

3924

3925

3926

3927

3928

3929

3930

3931

3932

3933

3934

3935

3936

3937

3938

3939

3940

3941

3942

3943

3944

3945

3946

3947

3948

3949

3950

3951

3952

3953

3954

3955

3956

3957

3958

3959

3960

3961

3962

3963

3964

3965

3966

3967

3968

3969

3970

3971

3972

3973

3974

3975

3976

3977

3978

3979

3980

3981

3982

3983

3984

3985

3986

3987

3988

3989

3990

3991

3992

3993

3994

3995

3996

3997

3998

3999

4000

4001

4002

4003

4004

4005

4006

4007

4008

4009

4010

4011

""" 

 

Numerical python functions written for compatibility with MATLAB 

commands with the same names. 

 

MATLAB compatible functions 

--------------------------- 

 

:func:`cohere` 

Coherence (normalized cross spectral density) 

 

:func:`csd` 

Cross spectral density using Welch's average periodogram 

 

:func:`detrend` 

Remove the mean or best fit line from an array 

 

:func:`find` 

Return the indices where some condition is true; 

numpy.nonzero is similar but more general. 

 

:func:`griddata` 

Interpolate irregularly distributed data to a 

regular grid. 

 

:func:`prctile` 

Find the percentiles of a sequence 

 

:func:`prepca` 

Principal Component Analysis 

 

:func:`psd` 

Power spectral density using Welch's average periodogram 

 

:func:`rk4` 

A 4th order runge kutta integrator for 1D or ND systems 

 

:func:`specgram` 

Spectrogram (spectrum over segments of time) 

 

Miscellaneous functions 

----------------------- 

 

Functions that don't exist in MATLAB, but are useful anyway: 

 

:func:`cohere_pairs` 

Coherence over all pairs. This is not a MATLAB function, but we 

compute coherence a lot in my lab, and we compute it for a lot of 

pairs. This function is optimized to do this efficiently by 

caching the direct FFTs. 

 

:func:`rk4` 

A 4th order Runge-Kutta ODE integrator in case you ever find 

yourself stranded without scipy (and the far superior 

scipy.integrate tools) 

 

:func:`contiguous_regions` 

Return the indices of the regions spanned by some logical mask 

 

:func:`cross_from_below` 

Return the indices where a 1D array crosses a threshold from below 

 

:func:`cross_from_above` 

Return the indices where a 1D array crosses a threshold from above 

 

:func:`complex_spectrum` 

Return the complex-valued frequency spectrum of a signal 

 

:func:`magnitude_spectrum` 

Return the magnitude of the frequency spectrum of a signal 

 

:func:`angle_spectrum` 

Return the angle (wrapped phase) of the frequency spectrum of a signal 

 

:func:`phase_spectrum` 

Return the phase (unwrapped angle) of the frequency spectrum of a signal 

 

:func:`detrend_mean` 

Remove the mean from a line. 

 

:func:`demean` 

Remove the mean from a line. This function is the same as 

:func:`detrend_mean` except for the default *axis*. 

 

:func:`detrend_linear` 

Remove the best fit line from a line. 

 

:func:`detrend_none` 

Return the original line. 

 

:func:`stride_windows` 

Get all windows in an array in a memory-efficient manner 

 

:func:`stride_repeat` 

Repeat an array in a memory-efficient manner 

 

:func:`apply_window` 

Apply a window along a given axis 

 

 

record array helper functions 

----------------------------- 

 

A collection of helper methods for numpyrecord arrays 

 

.. _htmlonly: 

 

See :ref:`misc-examples-index` 

 

:func:`rec2txt` 

Pretty print a record array 

 

:func:`rec2csv` 

Store record array in CSV file 

 

:func:`csv2rec` 

Import record array from CSV file with type inspection 

 

:func:`rec_append_fields` 

Adds field(s)/array(s) to record array 

 

:func:`rec_drop_fields` 

Drop fields from record array 

 

:func:`rec_join` 

Join two record arrays on sequence of fields 

 

:func:`recs_join` 

A simple join of multiple recarrays using a single column as a key 

 

:func:`rec_groupby` 

Summarize data by groups (similar to SQL GROUP BY) 

 

:func:`rec_summarize` 

Helper code to filter rec array fields into new fields 

 

For the rec viewer functions(e rec2csv), there are a bunch of Format 

objects you can pass into the functions that will do things like color 

negative values red, set percent formatting and scaling, etc. 

 

Example usage:: 

 

r = csv2rec('somefile.csv', checkrows=0) 

 

formatd = dict( 

weight = FormatFloat(2), 

change = FormatPercent(2), 

cost = FormatThousands(2), 

) 

 

 

rec2excel(r, 'test.xls', formatd=formatd) 

rec2csv(r, 'test.csv', formatd=formatd) 

 

""" 

 

import copy 

import csv 

import operator 

import os 

import warnings 

 

import numpy as np 

 

import matplotlib.cbook as cbook 

from matplotlib import docstring 

from matplotlib.path import Path 

import math 

 

 

@cbook.deprecated("2.2", alternative='numpy.logspace or numpy.geomspace') 

def logspace(xmin, xmax, N): 

''' 

Return N values logarithmically spaced between xmin and xmax. 

 

''' 

return np.exp(np.linspace(np.log(xmin), np.log(xmax), N)) 

 

 

def window_hanning(x): 

''' 

Return x times the hanning window of len(x). 

 

See Also 

-------- 

:func:`window_none` 

:func:`window_none` is another window algorithm. 

''' 

return np.hanning(len(x))*x 

 

 

def window_none(x): 

''' 

No window function; simply return x. 

 

See Also 

-------- 

:func:`window_hanning` 

:func:`window_hanning` is another window algorithm. 

''' 

return x 

 

 

def apply_window(x, window, axis=0, return_window=None): 

''' 

Apply the given window to the given 1D or 2D array along the given axis. 

 

Parameters 

---------- 

x : 1D or 2D array or sequence 

Array or sequence containing the data. 

 

window : function or array. 

Either a function to generate a window or an array with length 

*x*.shape[*axis*] 

 

axis : integer 

The axis over which to do the repetition. 

Must be 0 or 1. The default is 0 

 

return_window : bool 

If true, also return the 1D values of the window that was applied 

''' 

x = np.asarray(x) 

 

if x.ndim < 1 or x.ndim > 2: 

raise ValueError('only 1D or 2D arrays can be used') 

if axis+1 > x.ndim: 

raise ValueError('axis(=%s) out of bounds' % axis) 

 

xshape = list(x.shape) 

xshapetarg = xshape.pop(axis) 

 

if cbook.iterable(window): 

if len(window) != xshapetarg: 

raise ValueError('The len(window) must be the same as the shape ' 

'of x for the chosen axis') 

windowVals = window 

else: 

windowVals = window(np.ones(xshapetarg, dtype=x.dtype)) 

 

if x.ndim == 1: 

if return_window: 

return windowVals * x, windowVals 

else: 

return windowVals * x 

 

xshapeother = xshape.pop() 

 

otheraxis = (axis+1) % 2 

 

windowValsRep = stride_repeat(windowVals, xshapeother, axis=otheraxis) 

 

if return_window: 

return windowValsRep * x, windowVals 

else: 

return windowValsRep * x 

 

 

def detrend(x, key=None, axis=None): 

''' 

Return x with its trend removed. 

 

Parameters 

---------- 

x : array or sequence 

Array or sequence containing the data. 

 

key : [ 'default' | 'constant' | 'mean' | 'linear' | 'none'] or function 

Specifies the detrend algorithm to use. 'default' is 'mean', which is 

the same as :func:`detrend_mean`. 'constant' is the same. 'linear' is 

the same as :func:`detrend_linear`. 'none' is the same as 

:func:`detrend_none`. The default is 'mean'. See the corresponding 

functions for more details regarding the algorithms. Can also be a 

function that carries out the detrend operation. 

 

axis : integer 

The axis along which to do the detrending. 

 

See Also 

-------- 

:func:`detrend_mean` 

:func:`detrend_mean` implements the 'mean' algorithm. 

 

:func:`detrend_linear` 

:func:`detrend_linear` implements the 'linear' algorithm. 

 

:func:`detrend_none` 

:func:`detrend_none` implements the 'none' algorithm. 

''' 

if key is None or key in ['constant', 'mean', 'default']: 

return detrend(x, key=detrend_mean, axis=axis) 

elif key == 'linear': 

return detrend(x, key=detrend_linear, axis=axis) 

elif key == 'none': 

return detrend(x, key=detrend_none, axis=axis) 

elif isinstance(key, str): 

raise ValueError("Unknown value for key %s, must be one of: " 

"'default', 'constant', 'mean', " 

"'linear', or a function" % key) 

 

if not callable(key): 

raise ValueError("Unknown value for key %s, must be one of: " 

"'default', 'constant', 'mean', " 

"'linear', or a function" % key) 

 

x = np.asarray(x) 

 

if axis is not None and axis+1 > x.ndim: 

raise ValueError('axis(=%s) out of bounds' % axis) 

 

if (axis is None and x.ndim == 0) or (not axis and x.ndim == 1): 

return key(x) 

 

# try to use the 'axis' argument if the function supports it, 

# otherwise use apply_along_axis to do it 

try: 

return key(x, axis=axis) 

except TypeError: 

return np.apply_along_axis(key, axis=axis, arr=x) 

 

 

def demean(x, axis=0): 

''' 

Return x minus its mean along the specified axis. 

 

Parameters 

---------- 

x : array or sequence 

Array or sequence containing the data 

Can have any dimensionality 

 

axis : integer 

The axis along which to take the mean. See numpy.mean for a 

description of this argument. 

 

See Also 

-------- 

:func:`delinear` 

 

:func:`denone` 

:func:`delinear` and :func:`denone` are other detrend algorithms. 

 

:func:`detrend_mean` 

This function is the same as :func:`detrend_mean` except for the 

default *axis*. 

''' 

return detrend_mean(x, axis=axis) 

 

 

def detrend_mean(x, axis=None): 

''' 

Return x minus the mean(x). 

 

Parameters 

---------- 

x : array or sequence 

Array or sequence containing the data 

Can have any dimensionality 

 

axis : integer 

The axis along which to take the mean. See numpy.mean for a 

description of this argument. 

 

See Also 

-------- 

:func:`demean` 

This function is the same as :func:`demean` except for the default 

*axis*. 

 

:func:`detrend_linear` 

 

:func:`detrend_none` 

:func:`detrend_linear` and :func:`detrend_none` are other detrend 

algorithms. 

 

:func:`detrend` 

:func:`detrend` is a wrapper around all the detrend algorithms. 

''' 

x = np.asarray(x) 

 

if axis is not None and axis+1 > x.ndim: 

raise ValueError('axis(=%s) out of bounds' % axis) 

 

return x - x.mean(axis, keepdims=True) 

 

 

def detrend_none(x, axis=None): 

''' 

Return x: no detrending. 

 

Parameters 

---------- 

x : any object 

An object containing the data 

 

axis : integer 

This parameter is ignored. 

It is included for compatibility with detrend_mean 

 

See Also 

-------- 

:func:`denone` 

This function is the same as :func:`denone` except for the default 

*axis*, which has no effect. 

 

:func:`detrend_mean` 

 

:func:`detrend_linear` 

:func:`detrend_mean` and :func:`detrend_linear` are other detrend 

algorithms. 

 

:func:`detrend` 

:func:`detrend` is a wrapper around all the detrend algorithms. 

''' 

return x 

 

 

def detrend_linear(y): 

''' 

Return x minus best fit line; 'linear' detrending. 

 

Parameters 

---------- 

y : 0-D or 1-D array or sequence 

Array or sequence containing the data 

 

axis : integer 

The axis along which to take the mean. See numpy.mean for a 

description of this argument. 

 

See Also 

-------- 

:func:`delinear` 

This function is the same as :func:`delinear` except for the default 

*axis*. 

 

:func:`detrend_mean` 

 

:func:`detrend_none` 

:func:`detrend_mean` and :func:`detrend_none` are other detrend 

algorithms. 

 

:func:`detrend` 

:func:`detrend` is a wrapper around all the detrend algorithms. 

''' 

# This is faster than an algorithm based on linalg.lstsq. 

y = np.asarray(y) 

 

if y.ndim > 1: 

raise ValueError('y cannot have ndim > 1') 

 

# short-circuit 0-D array. 

if not y.ndim: 

return np.array(0., dtype=y.dtype) 

 

x = np.arange(y.size, dtype=float) 

 

C = np.cov(x, y, bias=1) 

b = C[0, 1]/C[0, 0] 

 

a = y.mean() - b*x.mean() 

return y - (b*x + a) 

 

 

def stride_windows(x, n, noverlap=None, axis=0): 

''' 

Get all windows of x with length n as a single array, 

using strides to avoid data duplication. 

 

.. warning:: 

 

It is not safe to write to the output array. Multiple 

elements may point to the same piece of memory, 

so modifying one value may change others. 

 

Parameters 

---------- 

x : 1D array or sequence 

Array or sequence containing the data. 

 

n : integer 

The number of data points in each window. 

 

noverlap : integer 

The overlap between adjacent windows. 

Default is 0 (no overlap) 

 

axis : integer 

The axis along which the windows will run. 

 

References 

---------- 

`stackoverflow: Rolling window for 1D arrays in Numpy? 

<http://stackoverflow.com/a/6811241>`_ 

`stackoverflow: Using strides for an efficient moving average filter 

<http://stackoverflow.com/a/4947453>`_ 

''' 

if noverlap is None: 

noverlap = 0 

 

if noverlap >= n: 

raise ValueError('noverlap must be less than n') 

if n < 1: 

raise ValueError('n cannot be less than 1') 

 

x = np.asarray(x) 

 

if x.ndim != 1: 

raise ValueError('only 1-dimensional arrays can be used') 

if n == 1 and noverlap == 0: 

if axis == 0: 

return x[np.newaxis] 

else: 

return x[np.newaxis].transpose() 

if n > x.size: 

raise ValueError('n cannot be greater than the length of x') 

 

# np.lib.stride_tricks.as_strided easily leads to memory corruption for 

# non integer shape and strides, i.e. noverlap or n. See #3845. 

noverlap = int(noverlap) 

n = int(n) 

 

step = n - noverlap 

if axis == 0: 

shape = (n, (x.shape[-1]-noverlap)//step) 

strides = (x.strides[0], step*x.strides[0]) 

else: 

shape = ((x.shape[-1]-noverlap)//step, n) 

strides = (step*x.strides[0], x.strides[0]) 

return np.lib.stride_tricks.as_strided(x, shape=shape, strides=strides) 

 

 

def stride_repeat(x, n, axis=0): 

''' 

Repeat the values in an array in a memory-efficient manner. Array x is 

stacked vertically n times. 

 

.. warning:: 

 

It is not safe to write to the output array. Multiple 

elements may point to the same piece of memory, so 

modifying one value may change others. 

 

Parameters 

---------- 

x : 1D array or sequence 

Array or sequence containing the data. 

 

n : integer 

The number of time to repeat the array. 

 

axis : integer 

The axis along which the data will run. 

 

References 

---------- 

`stackoverflow: Repeat NumPy array without replicating data? 

<http://stackoverflow.com/a/5568169>`_ 

''' 

if axis not in [0, 1]: 

raise ValueError('axis must be 0 or 1') 

x = np.asarray(x) 

if x.ndim != 1: 

raise ValueError('only 1-dimensional arrays can be used') 

 

if n == 1: 

if axis == 0: 

return np.atleast_2d(x) 

else: 

return np.atleast_2d(x).T 

if n < 1: 

raise ValueError('n cannot be less than 1') 

 

# np.lib.stride_tricks.as_strided easily leads to memory corruption for 

# non integer shape and strides, i.e. n. See #3845. 

n = int(n) 

 

if axis == 0: 

shape = (n, x.size) 

strides = (0, x.strides[0]) 

else: 

shape = (x.size, n) 

strides = (x.strides[0], 0) 

 

return np.lib.stride_tricks.as_strided(x, shape=shape, strides=strides) 

 

 

def _spectral_helper(x, y=None, NFFT=None, Fs=None, detrend_func=None, 

window=None, noverlap=None, pad_to=None, 

sides=None, scale_by_freq=None, mode=None): 

''' 

This is a helper function that implements the commonality between the 

psd, csd, spectrogram and complex, magnitude, angle, and phase spectrums. 

It is *NOT* meant to be used outside of mlab and may change at any time. 

''' 

if y is None: 

# if y is None use x for y 

same_data = True 

else: 

# The checks for if y is x are so that we can use the same function to 

# implement the core of psd(), csd(), and spectrogram() without doing 

# extra calculations. We return the unaveraged Pxy, freqs, and t. 

same_data = y is x 

 

if Fs is None: 

Fs = 2 

if noverlap is None: 

noverlap = 0 

if detrend_func is None: 

detrend_func = detrend_none 

if window is None: 

window = window_hanning 

 

# if NFFT is set to None use the whole signal 

if NFFT is None: 

NFFT = 256 

 

if mode is None or mode == 'default': 

mode = 'psd' 

elif mode not in ['psd', 'complex', 'magnitude', 'angle', 'phase']: 

raise ValueError("Unknown value for mode %s, must be one of: " 

"'default', 'psd', 'complex', " 

"'magnitude', 'angle', 'phase'" % mode) 

 

if not same_data and mode != 'psd': 

raise ValueError("x and y must be equal if mode is not 'psd'") 

 

# Make sure we're dealing with a numpy array. If y and x were the same 

# object to start with, keep them that way 

x = np.asarray(x) 

if not same_data: 

y = np.asarray(y) 

 

if sides is None or sides == 'default': 

if np.iscomplexobj(x): 

sides = 'twosided' 

else: 

sides = 'onesided' 

elif sides not in ['onesided', 'twosided']: 

raise ValueError("Unknown value for sides %s, must be one of: " 

"'default', 'onesided', or 'twosided'" % sides) 

 

# zero pad x and y up to NFFT if they are shorter than NFFT 

if len(x) < NFFT: 

n = len(x) 

x = np.resize(x, (NFFT,)) 

x[n:] = 0 

 

if not same_data and len(y) < NFFT: 

n = len(y) 

y = np.resize(y, (NFFT,)) 

y[n:] = 0 

 

if pad_to is None: 

pad_to = NFFT 

 

if mode != 'psd': 

scale_by_freq = False 

elif scale_by_freq is None: 

scale_by_freq = True 

 

# For real x, ignore the negative frequencies unless told otherwise 

if sides == 'twosided': 

numFreqs = pad_to 

if pad_to % 2: 

freqcenter = (pad_to - 1)//2 + 1 

else: 

freqcenter = pad_to//2 

scaling_factor = 1. 

elif sides == 'onesided': 

if pad_to % 2: 

numFreqs = (pad_to + 1)//2 

else: 

numFreqs = pad_to//2 + 1 

scaling_factor = 2. 

 

result = stride_windows(x, NFFT, noverlap, axis=0) 

result = detrend(result, detrend_func, axis=0) 

result, windowVals = apply_window(result, window, axis=0, 

return_window=True) 

result = np.fft.fft(result, n=pad_to, axis=0)[:numFreqs, :] 

freqs = np.fft.fftfreq(pad_to, 1/Fs)[:numFreqs] 

 

if not same_data: 

# if same_data is False, mode must be 'psd' 

resultY = stride_windows(y, NFFT, noverlap) 

resultY = detrend(resultY, detrend_func, axis=0) 

resultY = apply_window(resultY, window, axis=0) 

resultY = np.fft.fft(resultY, n=pad_to, axis=0)[:numFreqs, :] 

result = np.conj(result) * resultY 

elif mode == 'psd': 

result = np.conj(result) * result 

elif mode == 'magnitude': 

result = np.abs(result) / np.abs(windowVals).sum() 

elif mode == 'angle' or mode == 'phase': 

# we unwrap the phase later to handle the onesided vs. twosided case 

result = np.angle(result) 

elif mode == 'complex': 

result /= np.abs(windowVals).sum() 

 

if mode == 'psd': 

 

# Also include scaling factors for one-sided densities and dividing by 

# the sampling frequency, if desired. Scale everything, except the DC 

# component and the NFFT/2 component: 

 

# if we have a even number of frequencies, don't scale NFFT/2 

if not NFFT % 2: 

slc = slice(1, -1, None) 

# if we have an odd number, just don't scale DC 

else: 

slc = slice(1, None, None) 

 

result[slc] *= scaling_factor 

 

# MATLAB divides by the sampling frequency so that density function 

# has units of dB/Hz and can be integrated by the plotted frequency 

# values. Perform the same scaling here. 

if scale_by_freq: 

result /= Fs 

# Scale the spectrum by the norm of the window to compensate for 

# windowing loss; see Bendat & Piersol Sec 11.5.2. 

result /= (np.abs(windowVals)**2).sum() 

else: 

# In this case, preserve power in the segment, not amplitude 

result /= np.abs(windowVals).sum()**2 

 

t = np.arange(NFFT/2, len(x) - NFFT/2 + 1, NFFT - noverlap)/Fs 

 

if sides == 'twosided': 

# center the frequency range at zero 

freqs = np.concatenate((freqs[freqcenter:], freqs[:freqcenter])) 

result = np.concatenate((result[freqcenter:, :], 

result[:freqcenter, :]), 0) 

elif not pad_to % 2: 

# get the last value correctly, it is negative otherwise 

freqs[-1] *= -1 

 

# we unwrap the phase here to handle the onesided vs. twosided case 

if mode == 'phase': 

result = np.unwrap(result, axis=0) 

 

return result, freqs, t 

 

 

def _single_spectrum_helper(x, mode, Fs=None, window=None, pad_to=None, 

sides=None): 

''' 

This is a helper function that implements the commonality between the 

complex, magnitude, angle, and phase spectrums. 

It is *NOT* meant to be used outside of mlab and may change at any time. 

''' 

if mode is None or mode == 'psd' or mode == 'default': 

raise ValueError('_single_spectrum_helper does not work with %s mode' 

% mode) 

 

if pad_to is None: 

pad_to = len(x) 

 

spec, freqs, _ = _spectral_helper(x=x, y=None, NFFT=len(x), Fs=Fs, 

detrend_func=detrend_none, window=window, 

noverlap=0, pad_to=pad_to, 

sides=sides, 

scale_by_freq=False, 

mode=mode) 

if mode != 'complex': 

spec = spec.real 

 

if spec.ndim == 2 and spec.shape[1] == 1: 

spec = spec[:, 0] 

 

return spec, freqs 

 

 

# Split out these keyword docs so that they can be used elsewhere 

docstring.interpd.update(Spectral=cbook.dedent(""" 

Fs : scalar 

The sampling frequency (samples per time unit). It is used 

to calculate the Fourier frequencies, freqs, in cycles per time 

unit. The default value is 2. 

 

window : callable or ndarray 

A function or a vector of length *NFFT*. To create window 

vectors see :func:`window_hanning`, :func:`window_none`, 

:func:`numpy.blackman`, :func:`numpy.hamming`, 

:func:`numpy.bartlett`, :func:`scipy.signal`, 

:func:`scipy.signal.get_window`, etc. The default is 

:func:`window_hanning`. If a function is passed as the 

argument, it must take a data segment as an argument and 

return the windowed version of the segment. 

 

sides : {'default', 'onesided', 'twosided'} 

Specifies which sides of the spectrum to return. Default gives the 

default behavior, which returns one-sided for real data and both 

for complex data. 'onesided' forces the return of a one-sided 

spectrum, while 'twosided' forces two-sided. 

""")) 

 

 

docstring.interpd.update(Single_Spectrum=cbook.dedent(""" 

pad_to : int 

The number of points to which the data segment is padded when 

performing the FFT. While not increasing the actual resolution of 

the spectrum (the minimum distance between resolvable peaks), 

this can give more points in the plot, allowing for more 

detail. This corresponds to the *n* parameter in the call to fft(). 

The default is None, which sets *pad_to* equal to the length of the 

input signal (i.e. no padding). 

""")) 

 

 

docstring.interpd.update(PSD=cbook.dedent(""" 

pad_to : int 

The number of points to which the data segment is padded when 

performing the FFT. This can be different from *NFFT*, which 

specifies the number of data points used. While not increasing 

the actual resolution of the spectrum (the minimum distance between 

resolvable peaks), this can give more points in the plot, 

allowing for more detail. This corresponds to the *n* parameter 

in the call to fft(). The default is None, which sets *pad_to* 

equal to *NFFT* 

 

NFFT : int 

The number of data points used in each block for the FFT. 

A power 2 is most efficient. The default value is 256. 

This should *NOT* be used to get zero padding, or the scaling of the 

result will be incorrect. Use *pad_to* for this instead. 

 

detrend : {'default', 'constant', 'mean', 'linear', 'none'} or callable 

The function applied to each segment before fft-ing, 

designed to remove the mean or linear trend. Unlike in 

MATLAB, where the *detrend* parameter is a vector, in 

matplotlib is it a function. The :mod:`~matplotlib.mlab` 

module defines :func:`~matplotlib.mlab.detrend_none`, 

:func:`~matplotlib.mlab.detrend_mean`, and 

:func:`~matplotlib.mlab.detrend_linear`, but you can use 

a custom function as well. You can also use a string to choose 

one of the functions. 'default', 'constant', and 'mean' call 

:func:`~matplotlib.mlab.detrend_mean`. 'linear' calls 

:func:`~matplotlib.mlab.detrend_linear`. 'none' calls 

:func:`~matplotlib.mlab.detrend_none`. 

 

scale_by_freq : bool, optional 

Specifies whether the resulting density values should be scaled 

by the scaling frequency, which gives density in units of Hz^-1. 

This allows for integration over the returned frequency values. 

The default is True for MATLAB compatibility. 

""")) 

 

 

@docstring.dedent_interpd 

def psd(x, NFFT=None, Fs=None, detrend=None, window=None, 

noverlap=None, pad_to=None, sides=None, scale_by_freq=None): 

r""" 

Compute the power spectral density. 

 

Call signature:: 

 

psd(x, NFFT=256, Fs=2, detrend=mlab.detrend_none, 

window=mlab.window_hanning, noverlap=0, pad_to=None, 

sides='default', scale_by_freq=None) 

 

The power spectral density :math:`P_{xx}` by Welch's average 

periodogram method. The vector *x* is divided into *NFFT* length 

segments. Each segment is detrended by function *detrend* and 

windowed by function *window*. *noverlap* gives the length of 

the overlap between segments. The :math:`|\mathrm{fft}(i)|^2` 

of each segment :math:`i` are averaged to compute :math:`P_{xx}`. 

 

If len(*x*) < *NFFT*, it will be zero padded to *NFFT*. 

 

Parameters 

---------- 

x : 1-D array or sequence 

Array or sequence containing the data 

 

%(Spectral)s 

 

%(PSD)s 

 

noverlap : integer 

The number of points of overlap between segments. 

The default value is 0 (no overlap). 

 

Returns 

------- 

Pxx : 1-D array 

The values for the power spectrum `P_{xx}` (real valued) 

 

freqs : 1-D array 

The frequencies corresponding to the elements in *Pxx* 

 

References 

---------- 

Bendat & Piersol -- Random Data: Analysis and Measurement Procedures, John 

Wiley & Sons (1986) 

 

See Also 

-------- 

:func:`specgram` 

:func:`specgram` differs in the default overlap; in not returning the 

mean of the segment periodograms; and in returning the times of the 

segments. 

 

:func:`magnitude_spectrum` 

:func:`magnitude_spectrum` returns the magnitude spectrum. 

 

:func:`csd` 

:func:`csd` returns the spectral density between two signals. 

""" 

Pxx, freqs = csd(x=x, y=None, NFFT=NFFT, Fs=Fs, detrend=detrend, 

window=window, noverlap=noverlap, pad_to=pad_to, 

sides=sides, scale_by_freq=scale_by_freq) 

return Pxx.real, freqs 

 

 

@docstring.dedent_interpd 

def csd(x, y, NFFT=None, Fs=None, detrend=None, window=None, 

noverlap=None, pad_to=None, sides=None, scale_by_freq=None): 

""" 

Compute the cross-spectral density. 

 

Call signature:: 

 

csd(x, y, NFFT=256, Fs=2, detrend=mlab.detrend_none, 

window=mlab.window_hanning, noverlap=0, pad_to=None, 

sides='default', scale_by_freq=None) 

 

The cross spectral density :math:`P_{xy}` by Welch's average 

periodogram method. The vectors *x* and *y* are divided into 

*NFFT* length segments. Each segment is detrended by function 

*detrend* and windowed by function *window*. *noverlap* gives 

the length of the overlap between segments. The product of 

the direct FFTs of *x* and *y* are averaged over each segment 

to compute :math:`P_{xy}`, with a scaling to correct for power 

loss due to windowing. 

 

If len(*x*) < *NFFT* or len(*y*) < *NFFT*, they will be zero 

padded to *NFFT*. 

 

Parameters 

---------- 

x, y : 1-D arrays or sequences 

Arrays or sequences containing the data 

 

%(Spectral)s 

 

%(PSD)s 

 

noverlap : integer 

The number of points of overlap between segments. 

The default value is 0 (no overlap). 

 

Returns 

------- 

Pxy : 1-D array 

The values for the cross spectrum `P_{xy}` before scaling (real valued) 

 

freqs : 1-D array 

The frequencies corresponding to the elements in *Pxy* 

 

References 

---------- 

Bendat & Piersol -- Random Data: Analysis and Measurement Procedures, John 

Wiley & Sons (1986) 

 

See Also 

-------- 

:func:`psd` 

:func:`psd` is the equivalent to setting y=x. 

""" 

if NFFT is None: 

NFFT = 256 

Pxy, freqs, _ = _spectral_helper(x=x, y=y, NFFT=NFFT, Fs=Fs, 

detrend_func=detrend, window=window, 

noverlap=noverlap, pad_to=pad_to, 

sides=sides, scale_by_freq=scale_by_freq, 

mode='psd') 

 

if Pxy.ndim == 2: 

if Pxy.shape[1] > 1: 

Pxy = Pxy.mean(axis=1) 

else: 

Pxy = Pxy[:, 0] 

return Pxy, freqs 

 

 

@docstring.dedent_interpd 

def complex_spectrum(x, Fs=None, window=None, pad_to=None, 

sides=None): 

""" 

Compute the complex-valued frequency spectrum of *x*. Data is padded to a 

length of *pad_to* and the windowing function *window* is applied to the 

signal. 

 

Parameters 

---------- 

x : 1-D array or sequence 

Array or sequence containing the data 

 

%(Spectral)s 

 

%(Single_Spectrum)s 

 

Returns 

------- 

spectrum : 1-D array 

The values for the complex spectrum (complex valued) 

 

freqs : 1-D array 

The frequencies corresponding to the elements in *spectrum* 

 

See Also 

-------- 

:func:`magnitude_spectrum` 

:func:`magnitude_spectrum` returns the absolute value of this function. 

 

:func:`angle_spectrum` 

:func:`angle_spectrum` returns the angle of this function. 

 

:func:`phase_spectrum` 

:func:`phase_spectrum` returns the phase (unwrapped angle) of this 

function. 

 

:func:`specgram` 

:func:`specgram` can return the complex spectrum of segments within the 

signal. 

""" 

return _single_spectrum_helper(x=x, Fs=Fs, window=window, pad_to=pad_to, 

sides=sides, mode='complex') 

 

 

@docstring.dedent_interpd 

def magnitude_spectrum(x, Fs=None, window=None, pad_to=None, 

sides=None): 

""" 

Compute the magnitude (absolute value) of the frequency spectrum of 

*x*. Data is padded to a length of *pad_to* and the windowing function 

*window* is applied to the signal. 

 

Parameters 

---------- 

x : 1-D array or sequence 

Array or sequence containing the data 

 

%(Spectral)s 

 

%(Single_Spectrum)s 

 

Returns 

------- 

spectrum : 1-D array 

The values for the magnitude spectrum (real valued) 

 

freqs : 1-D array 

The frequencies corresponding to the elements in *spectrum* 

 

See Also 

-------- 

:func:`psd` 

:func:`psd` returns the power spectral density. 

 

:func:`complex_spectrum` 

This function returns the absolute value of :func:`complex_spectrum`. 

 

:func:`angle_spectrum` 

:func:`angle_spectrum` returns the angles of the corresponding 

frequencies. 

 

:func:`phase_spectrum` 

:func:`phase_spectrum` returns the phase (unwrapped angle) of the 

corresponding frequencies. 

 

:func:`specgram` 

:func:`specgram` can return the magnitude spectrum of segments within 

the signal. 

""" 

return _single_spectrum_helper(x=x, Fs=Fs, window=window, pad_to=pad_to, 

sides=sides, mode='magnitude') 

 

 

@docstring.dedent_interpd 

def angle_spectrum(x, Fs=None, window=None, pad_to=None, 

sides=None): 

""" 

Compute the angle of the frequency spectrum (wrapped phase spectrum) of 

*x*. Data is padded to a length of *pad_to* and the windowing function 

*window* is applied to the signal. 

 

Parameters 

---------- 

x : 1-D array or sequence 

Array or sequence containing the data 

 

%(Spectral)s 

 

%(Single_Spectrum)s 

 

Returns 

------- 

spectrum : 1-D array 

The values for the angle spectrum in radians (real valued) 

 

freqs : 1-D array 

The frequencies corresponding to the elements in *spectrum* 

 

See Also 

-------- 

:func:`complex_spectrum` 

This function returns the angle value of :func:`complex_spectrum`. 

 

:func:`magnitude_spectrum` 

:func:`angle_spectrum` returns the magnitudes of the corresponding 

frequencies. 

 

:func:`phase_spectrum` 

:func:`phase_spectrum` returns the unwrapped version of this function. 

 

:func:`specgram` 

:func:`specgram` can return the angle spectrum of segments within the 

signal. 

""" 

return _single_spectrum_helper(x=x, Fs=Fs, window=window, pad_to=pad_to, 

sides=sides, mode='angle') 

 

 

@docstring.dedent_interpd 

def phase_spectrum(x, Fs=None, window=None, pad_to=None, 

sides=None): 

""" 

Compute the phase of the frequency spectrum (unwrapped angle spectrum) of 

*x*. Data is padded to a length of *pad_to* and the windowing function 

*window* is applied to the signal. 

 

Parameters 

---------- 

x : 1-D array or sequence 

Array or sequence containing the data 

 

%(Spectral)s 

 

%(Single_Spectrum)s 

 

Returns 

------- 

spectrum : 1-D array 

The values for the phase spectrum in radians (real valued) 

 

freqs : 1-D array 

The frequencies corresponding to the elements in *spectrum* 

 

See Also 

-------- 

:func:`complex_spectrum` 

This function returns the angle value of :func:`complex_spectrum`. 

 

:func:`magnitude_spectrum` 

:func:`magnitude_spectrum` returns the magnitudes of the corresponding 

frequencies. 

 

:func:`angle_spectrum` 

:func:`angle_spectrum` returns the wrapped version of this function. 

 

:func:`specgram` 

:func:`specgram` can return the phase spectrum of segments within the 

signal. 

""" 

return _single_spectrum_helper(x=x, Fs=Fs, window=window, pad_to=pad_to, 

sides=sides, mode='phase') 

 

 

@docstring.dedent_interpd 

def specgram(x, NFFT=None, Fs=None, detrend=None, window=None, 

noverlap=None, pad_to=None, sides=None, scale_by_freq=None, 

mode=None): 

""" 

Compute a spectrogram. 

 

Compute and plot a spectrogram of data in x. Data are split into 

NFFT length segments and the spectrum of each section is 

computed. The windowing function window is applied to each 

segment, and the amount of overlap of each segment is 

specified with noverlap. 

 

Parameters 

---------- 

x : array_like 

1-D array or sequence. 

 

%(Spectral)s 

 

%(PSD)s 

 

noverlap : int, optional 

The number of points of overlap between blocks. The default 

value is 128. 

mode : str, optional 

What sort of spectrum to use, default is 'psd'. 

'psd' 

Returns the power spectral density. 

 

'complex' 

Returns the complex-valued frequency spectrum. 

 

'magnitude' 

Returns the magnitude spectrum. 

 

'angle' 

Returns the phase spectrum without unwrapping. 

 

'phase' 

Returns the phase spectrum with unwrapping. 

 

Returns 

------- 

spectrum : array_like 

2-D array, columns are the periodograms of successive segments. 

 

freqs : array_like 

1-D array, frequencies corresponding to the rows in *spectrum*. 

 

t : array_like 

1-D array, the times corresponding to midpoints of segments 

(i.e the columns in *spectrum*). 

 

See Also 

-------- 

psd : differs in the overlap and in the return values. 

complex_spectrum : similar, but with complex valued frequencies. 

magnitude_spectrum : similar single segment when mode is 'magnitude'. 

angle_spectrum : similar to single segment when mode is 'angle'. 

phase_spectrum : similar to single segment when mode is 'phase'. 

 

Notes 

----- 

detrend and scale_by_freq only apply when *mode* is set to 'psd'. 

 

""" 

if noverlap is None: 

noverlap = 128 # default in _spectral_helper() is noverlap = 0 

if NFFT is None: 

NFFT = 256 # same default as in _spectral_helper() 

if len(x) <= NFFT: 

warnings.warn("Only one segment is calculated since parameter NFFT " + 

"(=%d) >= signal length (=%d)." % (NFFT, len(x))) 

 

spec, freqs, t = _spectral_helper(x=x, y=None, NFFT=NFFT, Fs=Fs, 

detrend_func=detrend, window=window, 

noverlap=noverlap, pad_to=pad_to, 

sides=sides, 

scale_by_freq=scale_by_freq, 

mode=mode) 

 

if mode != 'complex': 

spec = spec.real # Needed since helper implements generically 

 

return spec, freqs, t 

 

 

_coh_error = """Coherence is calculated by averaging over *NFFT* 

length segments. Your signal is too short for your choice of *NFFT*. 

""" 

 

 

@docstring.dedent_interpd 

def cohere(x, y, NFFT=256, Fs=2, detrend=detrend_none, window=window_hanning, 

noverlap=0, pad_to=None, sides='default', scale_by_freq=None): 

""" 

The coherence between *x* and *y*. Coherence is the normalized 

cross spectral density: 

 

.. math:: 

 

C_{xy} = \\frac{|P_{xy}|^2}{P_{xx}P_{yy}} 

 

Parameters 

---------- 

x, y 

Array or sequence containing the data 

 

%(Spectral)s 

 

%(PSD)s 

 

noverlap : integer 

The number of points of overlap between blocks. The default value 

is 0 (no overlap). 

 

Returns 

------- 

The return value is the tuple (*Cxy*, *f*), where *f* are the 

frequencies of the coherence vector. For cohere, scaling the 

individual densities by the sampling frequency has no effect, 

since the factors cancel out. 

 

See Also 

-------- 

:func:`psd`, :func:`csd` : 

For information about the methods used to compute :math:`P_{xy}`, 

:math:`P_{xx}` and :math:`P_{yy}`. 

""" 

 

if len(x) < 2 * NFFT: 

raise ValueError(_coh_error) 

Pxx, f = psd(x, NFFT, Fs, detrend, window, noverlap, pad_to, sides, 

scale_by_freq) 

Pyy, f = psd(y, NFFT, Fs, detrend, window, noverlap, pad_to, sides, 

scale_by_freq) 

Pxy, f = csd(x, y, NFFT, Fs, detrend, window, noverlap, pad_to, sides, 

scale_by_freq) 

Cxy = np.abs(Pxy) ** 2 / (Pxx * Pyy) 

return Cxy, f 

 

 

@cbook.deprecated('2.2') 

def donothing_callback(*args): 

pass 

 

 

@cbook.deprecated('2.2', 'scipy.signal.coherence') 

def cohere_pairs(X, ij, NFFT=256, Fs=2, detrend=detrend_none, 

window=window_hanning, noverlap=0, 

preferSpeedOverMemory=True, 

progressCallback=donothing_callback, 

returnPxx=False): 

 

""" 

Compute the coherence and phase for all pairs *ij*, in *X*. 

 

*X* is a *numSamples* * *numCols* array 

 

*ij* is a list of tuples. Each tuple is a pair of indexes into 

the columns of X for which you want to compute coherence. For 

example, if *X* has 64 columns, and you want to compute all 

nonredundant pairs, define *ij* as:: 

 

ij = [] 

for i in range(64): 

for j in range(i+1,64): 

ij.append( (i,j) ) 

 

*preferSpeedOverMemory* is an optional bool. Defaults to true. If 

False, limits the caching by only making one, rather than two, 

complex cache arrays. This is useful if memory becomes critical. 

Even when *preferSpeedOverMemory* is False, :func:`cohere_pairs` 

will still give significant performance gains over calling 

:func:`cohere` for each pair, and will use subtantially less 

memory than if *preferSpeedOverMemory* is True. In my tests with 

a 43000,64 array over all nonredundant pairs, 

*preferSpeedOverMemory* = True delivered a 33% performance boost 

on a 1.7GHZ Athlon with 512MB RAM compared with 

*preferSpeedOverMemory* = False. But both solutions were more 

than 10x faster than naively crunching all possible pairs through 

:func:`cohere`. 

 

Returns 

------- 

Cxy : dictionary of (*i*, *j*) tuples -> coherence vector for 

that pair. i.e., ``Cxy[(i,j) = cohere(X[:,i], X[:,j])``. 

Number of dictionary keys is ``len(ij)``. 

 

Phase : dictionary of phases of the cross spectral density at 

each frequency for each pair. Keys are (*i*, *j*). 

 

freqs : vector of frequencies, equal in length to either the 

coherence or phase vectors for any (*i*, *j*) key. 

 

e.g., to make a coherence Bode plot:: 

 

subplot(211) 

plot( freqs, Cxy[(12,19)]) 

subplot(212) 

plot( freqs, Phase[(12,19)]) 

 

For a large number of pairs, :func:`cohere_pairs` can be much more 

efficient than just calling :func:`cohere` for each pair, because 

it caches most of the intensive computations. If :math:`N` is the 

number of pairs, this function is :math:`O(N)` for most of the 

heavy lifting, whereas calling cohere for each pair is 

:math:`O(N^2)`. However, because of the caching, it is also more 

memory intensive, making 2 additional complex arrays with 

approximately the same number of elements as *X*. 

 

See :file:`test/cohere_pairs_test.py` in the src tree for an 

example script that shows that this :func:`cohere_pairs` and 

:func:`cohere` give the same results for a given pair. 

 

See Also 

-------- 

:func:`psd` 

For information about the methods used to compute :math:`P_{xy}`, 

:math:`P_{xx}` and :math:`P_{yy}`. 

""" 

numRows, numCols = X.shape 

 

# zero pad if X is too short 

if numRows < NFFT: 

tmp = X 

X = np.zeros((NFFT, numCols), X.dtype) 

X[:numRows, :] = tmp 

del tmp 

 

numRows, numCols = X.shape 

# get all the columns of X that we are interested in by checking 

# the ij tuples 

allColumns = set() 

for i, j in ij: 

allColumns.add(i) 

allColumns.add(j) 

Ncols = len(allColumns) 

 

# for real X, ignore the negative frequencies 

if np.iscomplexobj(X): 

numFreqs = NFFT 

else: 

numFreqs = NFFT//2+1 

 

# cache the FFT of every windowed, detrended NFFT length segment 

# of every channel. If preferSpeedOverMemory, cache the conjugate 

# as well 

if cbook.iterable(window): 

if len(window) != NFFT: 

raise ValueError("The length of the window must be equal to NFFT") 

windowVals = window 

else: 

windowVals = window(np.ones(NFFT, X.dtype)) 

ind = list(range(0, numRows-NFFT+1, NFFT-noverlap)) 

numSlices = len(ind) 

FFTSlices = {} 

FFTConjSlices = {} 

Pxx = {} 

slices = range(numSlices) 

normVal = np.linalg.norm(windowVals)**2 

for iCol in allColumns: 

progressCallback(i/Ncols, 'Cacheing FFTs') 

Slices = np.zeros((numSlices, numFreqs), dtype=np.complex_) 

for iSlice in slices: 

thisSlice = X[ind[iSlice]:ind[iSlice]+NFFT, iCol] 

thisSlice = windowVals*detrend(thisSlice) 

Slices[iSlice, :] = np.fft.fft(thisSlice)[:numFreqs] 

 

FFTSlices[iCol] = Slices 

if preferSpeedOverMemory: 

FFTConjSlices[iCol] = np.conj(Slices) 

Pxx[iCol] = np.divide(np.mean(abs(Slices)**2, axis=0), normVal) 

del Slices, ind, windowVals 

 

# compute the coherences and phases for all pairs using the 

# cached FFTs 

Cxy = {} 

Phase = {} 

count = 0 

N = len(ij) 

for i, j in ij: 

count += 1 

if count % 10 == 0: 

progressCallback(count/N, 'Computing coherences') 

 

if preferSpeedOverMemory: 

Pxy = FFTSlices[i] * FFTConjSlices[j] 

else: 

Pxy = FFTSlices[i] * np.conj(FFTSlices[j]) 

if numSlices > 1: 

Pxy = np.mean(Pxy, axis=0) 

# Pxy = np.divide(Pxy, normVal) 

Pxy /= normVal 

# Cxy[(i,j)] = np.divide(np.absolute(Pxy)**2, Pxx[i]*Pxx[j]) 

Cxy[i, j] = abs(Pxy)**2 / (Pxx[i]*Pxx[j]) 

Phase[i, j] = np.arctan2(Pxy.imag, Pxy.real) 

 

freqs = Fs/NFFT*np.arange(numFreqs) 

if returnPxx: 

return Cxy, Phase, freqs, Pxx 

else: 

return Cxy, Phase, freqs 

 

 

@cbook.deprecated('2.2', 'scipy.stats.entropy') 

def entropy(y, bins): 

r""" 

Return the entropy of the data in *y* in units of nat. 

 

.. math:: 

 

-\sum p_i \ln(p_i) 

 

where :math:`p_i` is the probability of observing *y* in the 

:math:`i^{th}` bin of *bins*. *bins* can be a number of bins or a 

range of bins; see :func:`numpy.histogram`. 

 

Compare *S* with analytic calculation for a Gaussian:: 

 

x = mu + sigma * randn(200000) 

Sanalytic = 0.5 * ( 1.0 + log(2*pi*sigma**2.0) ) 

""" 

n, bins = np.histogram(y, bins) 

n = n.astype(float) 

 

n = np.take(n, np.nonzero(n)[0]) # get the positive 

 

p = np.divide(n, len(y)) 

 

delta = bins[1] - bins[0] 

S = -1.0 * np.sum(p * np.log(p)) + np.log(delta) 

return S 

 

 

@cbook.deprecated('2.2', 'scipy.stats.norm.pdf') 

def normpdf(x, *args): 

"Return the normal pdf evaluated at *x*; args provides *mu*, *sigma*" 

mu, sigma = args 

return 1./(np.sqrt(2*np.pi)*sigma)*np.exp(-0.5 * (1./sigma*(x - mu))**2) 

 

 

@cbook.deprecated('2.2') 

def find(condition): 

"Return the indices where ravel(condition) is true" 

res, = np.nonzero(np.ravel(condition)) 

return res 

 

 

@cbook.deprecated('2.2') 

def longest_contiguous_ones(x): 

""" 

Return the indices of the longest stretch of contiguous ones in *x*, 

assuming *x* is a vector of zeros and ones. If there are two 

equally long stretches, pick the first. 

""" 

x = np.ravel(x) 

if len(x) == 0: 

return np.array([]) 

 

ind = (x == 0).nonzero()[0] 

if len(ind) == 0: 

return np.arange(len(x)) 

if len(ind) == len(x): 

return np.array([]) 

 

y = np.zeros((len(x)+2,), x.dtype) 

y[1:-1] = x 

dif = np.diff(y) 

up = (dif == 1).nonzero()[0] 

dn = (dif == -1).nonzero()[0] 

i = (dn-up == max(dn - up)).nonzero()[0][0] 

ind = np.arange(up[i], dn[i]) 

 

return ind 

 

 

@cbook.deprecated('2.2') 

def longest_ones(x): 

'''alias for longest_contiguous_ones''' 

return longest_contiguous_ones(x) 

 

 

@cbook.deprecated('2.2') 

class PCA(object): 

def __init__(self, a, standardize=True): 

""" 

compute the SVD of a and store data for PCA. Use project to 

project the data onto a reduced set of dimensions 

 

Parameters 

---------- 

a : np.ndarray 

A numobservations x numdims array 

standardize : bool 

True if input data are to be standardized. If False, only centering 

will be carried out. 

 

Attributes 

---------- 

a 

A centered unit sigma version of input ``a``. 

 

numrows, numcols 

The dimensions of ``a``. 

 

mu 

A numdims array of means of ``a``. This is the vector that points 

to the origin of PCA space. 

 

sigma 

A numdims array of standard deviation of ``a``. 

 

fracs 

The proportion of variance of each of the principal components. 

 

s 

The actual eigenvalues of the decomposition. 

 

Wt 

The weight vector for projecting a numdims point or array into 

PCA space. 

 

Y 

A projected into PCA space. 

 

Notes 

----- 

The factor loadings are in the ``Wt`` factor, i.e., the factor loadings 

for the first principal component are given by ``Wt[0]``. This row is 

also the first eigenvector. 

 

""" 

n, m = a.shape 

if n < m: 

raise RuntimeError('we assume data in a is organized with ' 

'numrows>numcols') 

 

self.numrows, self.numcols = n, m 

self.mu = a.mean(axis=0) 

self.sigma = a.std(axis=0) 

self.standardize = standardize 

 

a = self.center(a) 

 

self.a = a 

 

U, s, Vh = np.linalg.svd(a, full_matrices=False) 

 

# Note: .H indicates the conjugate transposed / Hermitian. 

 

# The SVD is commonly written as a = U s V.H. 

# If U is a unitary matrix, it means that it satisfies U.H = inv(U). 

 

# The rows of Vh are the eigenvectors of a.H a. 

# The columns of U are the eigenvectors of a a.H. 

# For row i in Vh and column i in U, the corresponding eigenvalue is 

# s[i]**2. 

 

self.Wt = Vh 

 

# save the transposed coordinates 

Y = np.dot(Vh, a.T).T 

self.Y = Y 

 

# save the eigenvalues 

self.s = s**2 

 

# and now the contribution of the individual components 

vars = self.s / len(s) 

self.fracs = vars/vars.sum() 

 

def project(self, x, minfrac=0.): 

''' 

project x onto the principle axes, dropping any axes where fraction 

of variance<minfrac 

''' 

x = np.asarray(x) 

if x.shape[-1] != self.numcols: 

raise ValueError('Expected an array with dims[-1]==%d' % 

self.numcols) 

Y = np.dot(self.Wt, self.center(x).T).T 

mask = self.fracs >= minfrac 

if x.ndim == 2: 

Yreduced = Y[:, mask] 

else: 

Yreduced = Y[mask] 

return Yreduced 

 

def center(self, x): 

''' 

center and optionally standardize the data using the mean and sigma 

from training set a 

''' 

if self.standardize: 

return (x - self.mu)/self.sigma 

else: 

return (x - self.mu) 

 

@staticmethod 

def _get_colinear(): 

c0 = np.array([ 

0.19294738, 0.6202667, 0.45962655, 0.07608613, 0.135818, 

0.83580842, 0.07218851, 0.48318321, 0.84472463, 0.18348462, 

0.81585306, 0.96923926, 0.12835919, 0.35075355, 0.15807861, 

0.837437, 0.10824303, 0.1723387, 0.43926494, 0.83705486]) 

 

c1 = np.array([ 

-1.17705601, -0.513883, -0.26614584, 0.88067144, 1.00474954, 

-1.1616545, 0.0266109, 0.38227157, 1.80489433, 0.21472396, 

-1.41920399, -2.08158544, -0.10559009, 1.68999268, 0.34847107, 

-0.4685737, 1.23980423, -0.14638744, -0.35907697, 0.22442616]) 

 

c2 = c0 + 2*c1 

c3 = -3*c0 + 4*c1 

a = np.array([c3, c0, c1, c2]).T 

return a 

 

 

@cbook.deprecated('2.2', 'numpy.percentile') 

def prctile(x, p=(0.0, 25.0, 50.0, 75.0, 100.0)): 

""" 

Return the percentiles of *x*. *p* can either be a sequence of 

percentile values or a scalar. If *p* is a sequence, the ith 

element of the return sequence is the *p*(i)-th percentile of *x*. 

If *p* is a scalar, the largest value of *x* less than or equal to 

the *p* percentage point in the sequence is returned. 

""" 

 

# This implementation derived from scipy.stats.scoreatpercentile 

def _interpolate(a, b, fraction): 

"""Returns the point at the given fraction between a and b, where 

'fraction' must be between 0 and 1. 

""" 

return a + (b - a) * fraction 

 

per = np.array(p) 

values = np.sort(x, axis=None) 

 

idxs = per / 100 * (values.shape[0] - 1) 

ai = idxs.astype(int) 

bi = ai + 1 

frac = idxs % 1 

 

# handle cases where attempting to interpolate past last index 

cond = bi >= len(values) 

if per.ndim: 

ai[cond] -= 1 

bi[cond] -= 1 

frac[cond] += 1 

else: 

if cond: 

ai -= 1 

bi -= 1 

frac += 1 

 

return _interpolate(values[ai], values[bi], frac) 

 

 

@cbook.deprecated('2.2') 

def prctile_rank(x, p): 

""" 

Return the rank for each element in *x*, return the rank 

0..len(*p*). e.g., if *p* = (25, 50, 75), the return value will be a 

len(*x*) array with values in [0,1,2,3] where 0 indicates the 

value is less than the 25th percentile, 1 indicates the value is 

>= the 25th and < 50th percentile, ... and 3 indicates the value 

is above the 75th percentile cutoff. 

 

*p* is either an array of percentiles in [0..100] or a scalar which 

indicates how many quantiles of data you want ranked. 

""" 

 

if not cbook.iterable(p): 

p = np.arange(100.0/p, 100.0, 100.0/p) 

else: 

p = np.asarray(p) 

 

if p.max() <= 1 or p.min() < 0 or p.max() > 100: 

raise ValueError('percentiles should be in range 0..100, not 0..1') 

 

ptiles = prctile(x, p) 

return np.searchsorted(ptiles, x) 

 

 

@cbook.deprecated('2.2') 

def center_matrix(M, dim=0): 

""" 

Return the matrix *M* with each row having zero mean and unit std. 

 

If *dim* = 1 operate on columns instead of rows. (*dim* is 

opposite to the numpy axis kwarg.) 

""" 

M = np.asarray(M, float) 

if dim: 

M = (M - M.mean(axis=0)) / M.std(axis=0) 

else: 

M = (M - M.mean(axis=1)[:, np.newaxis]) 

M = M / M.std(axis=1)[:, np.newaxis] 

return M 

 

 

@cbook.deprecated('2.2', 'scipy.integrate.ode') 

def rk4(derivs, y0, t): 

""" 

Integrate 1D or ND system of ODEs using 4-th order Runge-Kutta. 

This is a toy implementation which may be useful if you find 

yourself stranded on a system w/o scipy. Otherwise use 

:func:`scipy.integrate`. 

 

Parameters 

---------- 

y0 

initial state vector 

 

t 

sample times 

 

derivs 

returns the derivative of the system and has the 

signature ``dy = derivs(yi, ti)`` 

 

Examples 

-------- 

 

A 2D system:: 

 

def derivs6(x,t): 

d1 = x[0] + 2*x[1] 

d2 = -3*x[0] + 4*x[1] 

return (d1, d2) 

dt = 0.0005 

t = arange(0.0, 2.0, dt) 

y0 = (1,2) 

yout = rk4(derivs6, y0, t) 

 

A 1D system:: 

 

alpha = 2 

def derivs(x,t): 

return -alpha*x + exp(-t) 

 

y0 = 1 

yout = rk4(derivs, y0, t) 

 

If you have access to scipy, you should probably be using the 

scipy.integrate tools rather than this function. 

""" 

 

try: 

Ny = len(y0) 

except TypeError: 

yout = np.zeros((len(t),), float) 

else: 

yout = np.zeros((len(t), Ny), float) 

 

yout[0] = y0 

i = 0 

 

for i in np.arange(len(t)-1): 

 

thist = t[i] 

dt = t[i+1] - thist 

dt2 = dt/2.0 

y0 = yout[i] 

 

k1 = np.asarray(derivs(y0, thist)) 

k2 = np.asarray(derivs(y0 + dt2*k1, thist+dt2)) 

k3 = np.asarray(derivs(y0 + dt2*k2, thist+dt2)) 

k4 = np.asarray(derivs(y0 + dt*k3, thist+dt)) 

yout[i+1] = y0 + dt/6.0*(k1 + 2*k2 + 2*k3 + k4) 

return yout 

 

 

@cbook.deprecated('2.2') 

def bivariate_normal(X, Y, sigmax=1.0, sigmay=1.0, 

mux=0.0, muy=0.0, sigmaxy=0.0): 

""" 

Bivariate Gaussian distribution for equal shape *X*, *Y*. 

 

See `bivariate normal 

<http://mathworld.wolfram.com/BivariateNormalDistribution.html>`_ 

at mathworld. 

""" 

Xmu = X-mux 

Ymu = Y-muy 

 

rho = sigmaxy/(sigmax*sigmay) 

z = Xmu**2/sigmax**2 + Ymu**2/sigmay**2 - 2*rho*Xmu*Ymu/(sigmax*sigmay) 

denom = 2*np.pi*sigmax*sigmay*np.sqrt(1-rho**2) 

return np.exp(-z/(2*(1-rho**2))) / denom 

 

 

@cbook.deprecated('2.2') 

def get_xyz_where(Z, Cond): 

""" 

*Z* and *Cond* are *M* x *N* matrices. *Z* are data and *Cond* is 

a boolean matrix where some condition is satisfied. Return value 

is (*x*, *y*, *z*) where *x* and *y* are the indices into *Z* and 

*z* are the values of *Z* at those indices. *x*, *y*, and *z* are 

1D arrays. 

""" 

X, Y = np.indices(Z.shape) 

return X[Cond], Y[Cond], Z[Cond] 

 

 

@cbook.deprecated('2.2') 

def get_sparse_matrix(M, N, frac=0.1): 

""" 

Return a *M* x *N* sparse matrix with *frac* elements randomly 

filled. 

""" 

data = np.zeros((M, N))*0. 

for i in range(int(M*N*frac)): 

x = np.random.randint(0, M-1) 

y = np.random.randint(0, N-1) 

data[x, y] = np.random.rand() 

return data 

 

 

@cbook.deprecated('2.2', 'numpy.hypot') 

def dist(x, y): 

""" 

Return the distance between two points. 

""" 

d = x-y 

return np.sqrt(np.dot(d, d)) 

 

 

@cbook.deprecated('2.2') 

def dist_point_to_segment(p, s0, s1): 

""" 

Get the distance of a point to a segment. 

 

*p*, *s0*, *s1* are *xy* sequences 

 

This algorithm from 

http://geomalgorithms.com/a02-_lines.html 

""" 

p = np.asarray(p, float) 

s0 = np.asarray(s0, float) 

s1 = np.asarray(s1, float) 

v = s1 - s0 

w = p - s0 

 

c1 = np.dot(w, v) 

if c1 <= 0: 

return dist(p, s0) 

 

c2 = np.dot(v, v) 

if c2 <= c1: 

return dist(p, s1) 

 

b = c1 / c2 

pb = s0 + b * v 

return dist(p, pb) 

 

 

@cbook.deprecated('2.2') 

def segments_intersect(s1, s2): 

""" 

Return *True* if *s1* and *s2* intersect. 

*s1* and *s2* are defined as:: 

 

s1: (x1, y1), (x2, y2) 

s2: (x3, y3), (x4, y4) 

""" 

(x1, y1), (x2, y2) = s1 

(x3, y3), (x4, y4) = s2 

 

den = ((y4-y3) * (x2-x1)) - ((x4-x3)*(y2-y1)) 

 

n1 = ((x4-x3) * (y1-y3)) - ((y4-y3)*(x1-x3)) 

n2 = ((x2-x1) * (y1-y3)) - ((y2-y1)*(x1-x3)) 

 

if den == 0: 

# lines parallel 

return False 

 

u1 = n1/den 

u2 = n2/den 

 

return 0.0 <= u1 <= 1.0 and 0.0 <= u2 <= 1.0 

 

 

@cbook.deprecated('2.2') 

def fftsurr(x, detrend=detrend_none, window=window_none): 

""" 

Compute an FFT phase randomized surrogate of *x*. 

""" 

if cbook.iterable(window): 

x = window*detrend(x) 

else: 

x = window(detrend(x)) 

z = np.fft.fft(x) 

a = 2.*np.pi*1j 

phase = a * np.random.rand(len(x)) 

z = z*np.exp(phase) 

return np.fft.ifft(z).real 

 

 

@cbook.deprecated('2.2') 

def movavg(x, n): 

""" 

Compute the len(*n*) moving average of *x*. 

""" 

w = np.empty((n,), dtype=float) 

w[:] = 1.0/n 

return np.convolve(x, w, mode='valid') 

 

 

# the following code was written and submitted by Fernando Perez 

# from the ipython numutils package under a BSD license 

# begin fperez functions 

 

""" 

A set of convenient utilities for numerical work. 

 

Most of this module requires numpy or is meant to be used with it. 

 

Copyright (c) 2001-2004, Fernando Perez. <Fernando.Perez@colorado.edu> 

All rights reserved. 

 

This license was generated from the BSD license template as found in: 

http://www.opensource.org/licenses/bsd-license.php 

 

Redistribution and use in source and binary forms, with or without 

modification, are permitted provided that the following conditions are met: 

 

* Redistributions of source code must retain the above copyright notice, 

this list of conditions and the following disclaimer. 

 

* Redistributions in binary form must reproduce the above copyright 

notice, this list of conditions and the following disclaimer in the 

documentation and/or other materials provided with the distribution. 

 

* Neither the name of the IPython project nor the names of its 

contributors may be used to endorse or promote products derived from 

this software without specific prior written permission. 

 

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 

AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 

IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 

DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE 

FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 

DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 

SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 

CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 

OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 

OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 

 

""" 

 

 

# ***************************************************************************** 

# Globals 

# **************************************************************************** 

# function definitions 

exp_safe_MIN = math.log(2.2250738585072014e-308) 

exp_safe_MAX = 1.7976931348623157e+308 

 

 

@cbook.deprecated("2.2", 'numpy.exp') 

def exp_safe(x): 

""" 

Compute exponentials which safely underflow to zero. 

 

Slow, but convenient to use. Note that numpy provides proper 

floating point exception handling with access to the underlying 

hardware. 

""" 

 

if type(x) is np.ndarray: 

return np.exp(np.clip(x, exp_safe_MIN, exp_safe_MAX)) 

else: 

return math.exp(x) 

 

 

@cbook.deprecated("2.2", alternative='numpy.array(list(map(...)))') 

def amap(fn, *args): 

""" 

amap(function, sequence[, sequence, ...]) -> array. 

 

Works like :func:`map`, but it returns an array. This is just a 

convenient shorthand for ``numpy.array(map(...))``. 

""" 

return np.array(list(map(fn, *args))) 

 

 

@cbook.deprecated("2.2") 

def rms_flat(a): 

""" 

Return the root mean square of all the elements of *a*, flattened out. 

""" 

return np.sqrt(np.mean(np.abs(a) ** 2)) 

 

 

@cbook.deprecated("2.2", alternative='numpy.linalg.norm(a, ord=1)') 

def l1norm(a): 

""" 

Return the *l1* norm of *a*, flattened out. 

 

Implemented as a separate function (not a call to :func:`norm` for speed). 

""" 

return np.sum(np.abs(a)) 

 

 

@cbook.deprecated("2.2", alternative='numpy.linalg.norm(a, ord=2)') 

def l2norm(a): 

""" 

Return the *l2* norm of *a*, flattened out. 

 

Implemented as a separate function (not a call to :func:`norm` for speed). 

""" 

return np.sqrt(np.sum(np.abs(a) ** 2)) 

 

 

@cbook.deprecated("2.2", alternative='numpy.linalg.norm(a.flat, ord=p)') 

def norm_flat(a, p=2): 

""" 

norm(a,p=2) -> l-p norm of a.flat 

 

Return the l-p norm of *a*, considered as a flat array. This is NOT a true 

matrix norm, since arrays of arbitrary rank are always flattened. 

 

*p* can be a number or the string 'Infinity' to get the L-infinity norm. 

""" 

# This function was being masked by a more general norm later in 

# the file. We may want to simply delete it. 

if p == 'Infinity': 

return np.max(np.abs(a)) 

else: 

return np.sum(np.abs(a) ** p) ** (1 / p) 

 

 

@cbook.deprecated("2.2", 'numpy.arange') 

def frange(xini, xfin=None, delta=None, **kw): 

""" 

frange([start,] stop[, step, keywords]) -> array of floats 

 

Return a numpy ndarray containing a progression of floats. Similar to 

:func:`numpy.arange`, but defaults to a closed interval. 

 

``frange(x0, x1)`` returns ``[x0, x0+1, x0+2, ..., x1]``; *start* 

defaults to 0, and the endpoint *is included*. This behavior is 

different from that of :func:`range` and 

:func:`numpy.arange`. This is deliberate, since :func:`frange` 

will probably be more useful for generating lists of points for 

function evaluation, and endpoints are often desired in this 

use. The usual behavior of :func:`range` can be obtained by 

setting the keyword *closed* = 0, in this case, :func:`frange` 

basically becomes :func:numpy.arange`. 

 

When *step* is given, it specifies the increment (or 

decrement). All arguments can be floating point numbers. 

 

``frange(x0,x1,d)`` returns ``[x0,x0+d,x0+2d,...,xfin]`` where 

*xfin* <= *x1*. 

 

:func:`frange` can also be called with the keyword *npts*. This 

sets the number of points the list should contain (and overrides 

the value *step* might have been given). :func:`numpy.arange` 

doesn't offer this option. 

 

Examples:: 

 

>>> frange(3) 

array([ 0., 1., 2., 3.]) 

>>> frange(3,closed=0) 

array([ 0., 1., 2.]) 

>>> frange(1,6,2) 

array([1, 3, 5]) or 1,3,5,7, depending on floating point vagueries 

>>> frange(1,6.5,npts=5) 

array([ 1. , 2.375, 3.75 , 5.125, 6.5 ]) 

""" 

 

# defaults 

kw.setdefault('closed', 1) 

endpoint = kw['closed'] != 0 

 

# funny logic to allow the *first* argument to be optional (like range()) 

# This was modified with a simpler version from a similar frange() found 

# at http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/66472 

if xfin is None: 

xfin = xini + 0.0 

xini = 0.0 

 

if delta is None: 

delta = 1.0 

 

# compute # of points, spacing and return final list 

try: 

npts = kw['npts'] 

delta = (xfin-xini) / (npts-endpoint) 

except KeyError: 

npts = int(np.round((xfin-xini)/delta)) + endpoint 

# round finds the nearest, so the endpoint can be up to 

# delta/2 larger than xfin. 

 

return np.arange(npts)*delta+xini 

# end frange() 

 

 

@cbook.deprecated("2.2", 'numpy.identity') 

def identity(n, rank=2, dtype='l', typecode=None): 

""" 

Returns the identity matrix of shape (*n*, *n*, ..., *n*) (rank *r*). 

 

For ranks higher than 2, this object is simply a multi-index Kronecker 

delta:: 

 

/ 1 if i0=i1=...=iR, 

id[i0,i1,...,iR] = -| 

\\ 0 otherwise. 

 

Optionally a *dtype* (or typecode) may be given (it defaults to 'l'). 

 

Since rank defaults to 2, this function behaves in the default case (when 

only *n* is given) like ``numpy.identity(n)`` -- but surprisingly, it is 

much faster. 

""" 

if typecode is not None: 

dtype = typecode 

iden = np.zeros((n,)*rank, dtype) 

for i in range(n): 

idx = (i,)*rank 

iden[idx] = 1 

return iden 

 

 

@cbook.deprecated("2.2") 

def base_repr(number, base=2, padding=0): 

""" 

Return the representation of a *number* in any given *base*. 

""" 

chars = '0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ' 

if number < base: 

return (padding - 1) * chars[0] + chars[int(number)] 

max_exponent = int(math.log(number)/math.log(base)) 

max_power = int(base) ** max_exponent 

lead_digit = int(number/max_power) 

return (chars[lead_digit] + 

base_repr(number - max_power * lead_digit, base, 

max(padding - 1, max_exponent))) 

 

 

@cbook.deprecated("2.2") 

def binary_repr(number, max_length=1025): 

""" 

Return the binary representation of the input *number* as a 

string. 

 

This is more efficient than using :func:`base_repr` with base 2. 

 

Increase the value of max_length for very large numbers. Note that 

on 32-bit machines, 2**1023 is the largest integer power of 2 

which can be converted to a Python float. 

""" 

 

# assert number < 2L << max_length 

shifts = map(operator.rshift, max_length * [number], 

range(max_length - 1, -1, -1)) 

digits = list(map(operator.mod, shifts, max_length * [2])) 

if not digits.count(1): 

return 0 

digits = digits[digits.index(1):] 

return ''.join(map(repr, digits)).replace('L', '') 

 

 

@cbook.deprecated("2.2", 'numpy.log2') 

def log2(x, ln2=math.log(2.0)): 

""" 

Return the log(*x*) in base 2. 

 

This is a _slow_ function but which is guaranteed to return the correct 

integer value if the input is an integer exact power of 2. 

""" 

try: 

bin_n = binary_repr(x)[1:] 

except (AssertionError, TypeError): 

return math.log(x)/ln2 

else: 

if '1' in bin_n: 

return math.log(x)/ln2 

else: 

return len(bin_n) 

 

 

@cbook.deprecated("2.2") 

def ispower2(n): 

""" 

Returns the log base 2 of *n* if *n* is a power of 2, zero otherwise. 

 

Note the potential ambiguity if *n* == 1: 2**0 == 1, interpret accordingly. 

""" 

 

bin_n = binary_repr(n)[1:] 

if '1' in bin_n: 

return 0 

else: 

return len(bin_n) 

 

 

@cbook.deprecated("2.2") 

def isvector(X): 

""" 

Like the MATLAB function with the same name, returns *True* 

if the supplied numpy array or matrix *X* looks like a vector, 

meaning it has a one non-singleton axis (i.e., it can have 

multiple axes, but all must have length 1, except for one of 

them). 

 

If you just want to see if the array has 1 axis, use X.ndim == 1. 

""" 

return np.prod(X.shape) == np.max(X.shape) 

 

# end fperez numutils code 

 

 

# helpers for loading, saving, manipulating and viewing numpy record arrays 

@cbook.deprecated("2.2", 'numpy.isnan') 

def safe_isnan(x): 

':func:`numpy.isnan` for arbitrary types' 

if isinstance(x, str): 

return False 

try: 

b = np.isnan(x) 

except NotImplementedError: 

return False 

except TypeError: 

return False 

else: 

return b 

 

 

@cbook.deprecated("2.2", 'numpy.isinf') 

def safe_isinf(x): 

':func:`numpy.isinf` for arbitrary types' 

if isinstance(x, str): 

return False 

try: 

b = np.isinf(x) 

except NotImplementedError: 

return False 

except TypeError: 

return False 

else: 

return b 

 

 

@cbook.deprecated("2.2") 

def rec_append_fields(rec, names, arrs, dtypes=None): 

""" 

Return a new record array with field names populated with data 

from arrays in *arrs*. If appending a single field, then *names*, 

*arrs* and *dtypes* do not have to be lists. They can just be the 

values themselves. 

""" 

if (not isinstance(names, str) and cbook.iterable(names) 

and len(names) and isinstance(names[0], str)): 

if len(names) != len(arrs): 

raise ValueError("number of arrays do not match number of names") 

else: # we have only 1 name and 1 array 

names = [names] 

arrs = [arrs] 

arrs = list(map(np.asarray, arrs)) 

if dtypes is None: 

dtypes = [a.dtype for a in arrs] 

elif not cbook.iterable(dtypes): 

dtypes = [dtypes] 

if len(arrs) != len(dtypes): 

if len(dtypes) == 1: 

dtypes = dtypes * len(arrs) 

else: 

raise ValueError("dtypes must be None, a single dtype or a list") 

old_dtypes = rec.dtype.descr 

newdtype = np.dtype(old_dtypes + list(zip(names, dtypes))) 

newrec = np.recarray(rec.shape, dtype=newdtype) 

for field in rec.dtype.fields: 

newrec[field] = rec[field] 

for name, arr in zip(names, arrs): 

newrec[name] = arr 

return newrec 

 

 

@cbook.deprecated("2.2") 

def rec_drop_fields(rec, names): 

""" 

Return a new numpy record array with fields in *names* dropped. 

""" 

 

names = set(names) 

 

newdtype = np.dtype([(name, rec.dtype[name]) for name in rec.dtype.names 

if name not in names]) 

 

newrec = np.recarray(rec.shape, dtype=newdtype) 

for field in newdtype.names: 

newrec[field] = rec[field] 

 

return newrec 

 

 

@cbook.deprecated("2.2") 

def rec_keep_fields(rec, names): 

""" 

Return a new numpy record array with only fields listed in names 

""" 

 

if isinstance(names, str): 

names = names.split(',') 

 

arrays = [] 

for name in names: 

arrays.append(rec[name]) 

 

return np.rec.fromarrays(arrays, names=names) 

 

 

@cbook.deprecated("2.2") 

def rec_groupby(r, groupby, stats): 

""" 

*r* is a numpy record array 

 

*groupby* is a sequence of record array attribute names that 

together form the grouping key. e.g., ('date', 'productcode') 

 

*stats* is a sequence of (*attr*, *func*, *outname*) tuples which 

will call ``x = func(attr)`` and assign *x* to the record array 

output with attribute *outname*. For example:: 

 

stats = ( ('sales', len, 'numsales'), ('sales', np.mean, 'avgsale') ) 

 

Return record array has *dtype* names for each attribute name in 

the *groupby* argument, with the associated group values, and 

for each outname name in the *stats* argument, with the associated 

stat summary output. 

""" 

# build a dictionary from groupby keys-> list of indices into r with 

# those keys 

rowd = {} 

for i, row in enumerate(r): 

key = tuple([row[attr] for attr in groupby]) 

rowd.setdefault(key, []).append(i) 

 

rows = [] 

# sort the output by groupby keys 

for key in sorted(rowd): 

row = list(key) 

# get the indices for this groupby key 

ind = rowd[key] 

thisr = r[ind] 

# call each stat function for this groupby slice 

row.extend([func(thisr[attr]) for attr, func, outname in stats]) 

rows.append(row) 

 

# build the output record array with groupby and outname attributes 

attrs, funcs, outnames = list(zip(*stats)) 

names = list(groupby) 

names.extend(outnames) 

return np.rec.fromrecords(rows, names=names) 

 

 

@cbook.deprecated("2.2") 

def rec_summarize(r, summaryfuncs): 

""" 

*r* is a numpy record array 

 

*summaryfuncs* is a list of (*attr*, *func*, *outname*) tuples 

which will apply *func* to the array *r*[attr] and assign the 

output to a new attribute name *outname*. The returned record 

array is identical to *r*, with extra arrays for each element in 

*summaryfuncs*. 

 

""" 

 

names = list(r.dtype.names) 

arrays = [r[name] for name in names] 

 

for attr, func, outname in summaryfuncs: 

names.append(outname) 

arrays.append(np.asarray(func(r[attr]))) 

 

return np.rec.fromarrays(arrays, names=names) 

 

 

@cbook.deprecated("2.2") 

def rec_join(key, r1, r2, jointype='inner', defaults=None, r1postfix='1', 

r2postfix='2'): 

""" 

Join record arrays *r1* and *r2* on *key*; *key* is a tuple of 

field names -- if *key* is a string it is assumed to be a single 

attribute name. If *r1* and *r2* have equal values on all the keys 

in the *key* tuple, then their fields will be merged into a new 

record array containing the intersection of the fields of *r1* and 

*r2*. 

 

*r1* (also *r2*) must not have any duplicate keys. 

 

The *jointype* keyword can be 'inner', 'outer', 'leftouter'. To 

do a rightouter join just reverse *r1* and *r2*. 

 

The *defaults* keyword is a dictionary filled with 

``{column_name:default_value}`` pairs. 

 

The keywords *r1postfix* and *r2postfix* are postfixed to column names 

(other than keys) that are both in *r1* and *r2*. 

""" 

 

if isinstance(key, str): 

key = (key, ) 

 

for name in key: 

if name not in r1.dtype.names: 

raise ValueError('r1 does not have key field %s' % name) 

if name not in r2.dtype.names: 

raise ValueError('r2 does not have key field %s' % name) 

 

def makekey(row): 

return tuple([row[name] for name in key]) 

 

r1d = {makekey(row): i for i, row in enumerate(r1)} 

r2d = {makekey(row): i for i, row in enumerate(r2)} 

 

r1keys = set(r1d) 

r2keys = set(r2d) 

 

common_keys = r1keys & r2keys 

 

r1ind = np.array([r1d[k] for k in common_keys]) 

r2ind = np.array([r2d[k] for k in common_keys]) 

 

common_len = len(common_keys) 

left_len = right_len = 0 

if jointype == "outer" or jointype == "leftouter": 

left_keys = r1keys.difference(r2keys) 

left_ind = np.array([r1d[k] for k in left_keys]) 

left_len = len(left_ind) 

if jointype == "outer": 

right_keys = r2keys.difference(r1keys) 

right_ind = np.array([r2d[k] for k in right_keys]) 

right_len = len(right_ind) 

 

def key_desc(name): 

''' 

if name is a string key, use the larger size of r1 or r2 before 

merging 

''' 

dt1 = r1.dtype[name] 

if dt1.type != np.string_: 

return (name, dt1.descr[0][1]) 

 

dt2 = r2.dtype[name] 

if dt1 != dt2: 

raise ValueError("The '{}' fields in arrays 'r1' and 'r2' must " 

"have the same dtype".format(name)) 

if dt1.num > dt2.num: 

return (name, dt1.descr[0][1]) 

else: 

return (name, dt2.descr[0][1]) 

 

keydesc = [key_desc(name) for name in key] 

 

def mapped_r1field(name): 

""" 

The column name in *newrec* that corresponds to the column in *r1*. 

""" 

if name in key or name not in r2.dtype.names: 

return name 

else: 

return name + r1postfix 

 

def mapped_r2field(name): 

""" 

The column name in *newrec* that corresponds to the column in *r2*. 

""" 

if name in key or name not in r1.dtype.names: 

return name 

else: 

return name + r2postfix 

 

r1desc = [(mapped_r1field(desc[0]), desc[1]) for desc in r1.dtype.descr 

if desc[0] not in key] 

r2desc = [(mapped_r2field(desc[0]), desc[1]) for desc in r2.dtype.descr 

if desc[0] not in key] 

all_dtypes = keydesc + r1desc + r2desc 

newdtype = np.dtype(all_dtypes) 

newrec = np.recarray((common_len + left_len + right_len,), dtype=newdtype) 

 

if defaults is not None: 

for thiskey in defaults: 

if thiskey not in newdtype.names: 

warnings.warn('rec_join defaults key="%s" not in new dtype ' 

'names "%s"' % (thiskey, newdtype.names)) 

 

for name in newdtype.names: 

dt = newdtype[name] 

if dt.kind in ('f', 'i'): 

newrec[name] = 0 

 

if jointype != 'inner' and defaults is not None: 

# fill in the defaults enmasse 

newrec_fields = list(newrec.dtype.fields) 

for k, v in defaults.items(): 

if k in newrec_fields: 

newrec[k] = v 

 

for field in r1.dtype.names: 

newfield = mapped_r1field(field) 

if common_len: 

newrec[newfield][:common_len] = r1[field][r1ind] 

if (jointype == "outer" or jointype == "leftouter") and left_len: 

newrec[newfield][common_len:(common_len+left_len)] = ( 

r1[field][left_ind] 

) 

 

for field in r2.dtype.names: 

newfield = mapped_r2field(field) 

if field not in key and common_len: 

newrec[newfield][:common_len] = r2[field][r2ind] 

if jointype == "outer" and right_len: 

newrec[newfield][-right_len:] = r2[field][right_ind] 

 

newrec.sort(order=key) 

 

return newrec 

 

 

@cbook.deprecated("2.2") 

def recs_join(key, name, recs, jointype='outer', missing=0., postfixes=None): 

""" 

Join a sequence of record arrays on single column key. 

 

This function only joins a single column of the multiple record arrays 

 

*key* 

is the column name that acts as a key 

 

*name* 

is the name of the column that we want to join 

 

*recs* 

is a list of record arrays to join 

 

*jointype* 

is a string 'inner' or 'outer' 

 

*missing* 

is what any missing field is replaced by 

 

*postfixes* 

if not None, a len recs sequence of postfixes 

 

returns a record array with columns [rowkey, name0, name1, ... namen-1]. 

or if postfixes [PF0, PF1, ..., PFN-1] are supplied, 

[rowkey, namePF0, namePF1, ... namePFN-1]. 

 

Example:: 

 

r = recs_join("date", "close", recs=[r0, r1], missing=0.) 

 

""" 

results = [] 

aligned_iters = cbook.align_iterators(operator.attrgetter(key), 

*[iter(r) for r in recs]) 

 

def extract(r): 

if r is None: 

return missing 

else: 

return r[name] 

 

if jointype == "outer": 

for rowkey, row in aligned_iters: 

results.append([rowkey] + list(map(extract, row))) 

elif jointype == "inner": 

for rowkey, row in aligned_iters: 

if None not in row: # throw out any Nones 

results.append([rowkey] + list(map(extract, row))) 

 

if postfixes is None: 

postfixes = ['%d' % i for i in range(len(recs))] 

names = ",".join([key] + ["%s%s" % (name, postfix) 

for postfix in postfixes]) 

return np.rec.fromrecords(results, names=names) 

 

 

@cbook.deprecated("2.2") 

def csv2rec(fname, comments='#', skiprows=0, checkrows=0, delimiter=',', 

converterd=None, names=None, missing='', missingd=None, 

use_mrecords=False, dayfirst=False, yearfirst=False): 

""" 

Load data from comma/space/tab delimited file in *fname* into a 

numpy record array and return the record array. 

 

If *names* is *None*, a header row is required to automatically 

assign the recarray names. The headers will be lower cased, 

spaces will be converted to underscores, and illegal attribute 

name characters removed. If *names* is not *None*, it is a 

sequence of names to use for the column names. In this case, it 

is assumed there is no header row. 

 

 

- *fname*: can be a filename or a file handle. Support for gzipped 

files is automatic, if the filename ends in '.gz' 

 

- *comments*: the character used to indicate the start of a comment 

in the file, or *None* to switch off the removal of comments 

 

- *skiprows*: is the number of rows from the top to skip 

 

- *checkrows*: is the number of rows to check to validate the column 

data type. When set to zero all rows are validated. 

 

- *converterd*: if not *None*, is a dictionary mapping column number or 

munged column name to a converter function. 

 

- *names*: if not None, is a list of header names. In this case, no 

header will be read from the file 

 

- *missingd* is a dictionary mapping munged column names to field values 

which signify that the field does not contain actual data and should 

be masked, e.g., '0000-00-00' or 'unused' 

 

- *missing*: a string whose value signals a missing field regardless of 

the column it appears in 

 

- *use_mrecords*: if True, return an mrecords.fromrecords record array if 

any of the data are missing 

 

- *dayfirst*: default is False so that MM-DD-YY has precedence over 

DD-MM-YY. See 

http://labix.org/python-dateutil#head-b95ce2094d189a89f80f5ae52a05b4ab7b41af47 

for further information. 

 

- *yearfirst*: default is False so that MM-DD-YY has precedence over 

YY-MM-DD. See 

http://labix.org/python-dateutil#head-b95ce2094d189a89f80f5ae52a05b4ab7b41af47 

for further information. 

 

If no rows are found, *None* is returned 

""" 

 

if converterd is None: 

converterd = dict() 

 

if missingd is None: 

missingd = {} 

 

import dateutil.parser 

import datetime 

 

fh = cbook.to_filehandle(fname) 

 

delimiter = str(delimiter) 

 

class FH: 

""" 

For space-delimited files, we want different behavior than 

comma or tab. Generally, we want multiple spaces to be 

treated as a single separator, whereas with comma and tab we 

want multiple commas to return multiple (empty) fields. The 

join/strip trick below effects this. 

""" 

def __init__(self, fh): 

self.fh = fh 

 

def close(self): 

self.fh.close() 

 

def seek(self, arg): 

self.fh.seek(arg) 

 

def fix(self, s): 

return ' '.join(s.split()) 

 

def __next__(self): 

return self.fix(next(self.fh)) 

 

def __iter__(self): 

for line in self.fh: 

yield self.fix(line) 

 

if delimiter == ' ': 

fh = FH(fh) 

 

reader = csv.reader(fh, delimiter=delimiter) 

 

def process_skiprows(reader): 

if skiprows: 

for i, row in enumerate(reader): 

if i >= (skiprows-1): 

break 

 

return fh, reader 

 

process_skiprows(reader) 

 

def ismissing(name, val): 

"Should the value val in column name be masked?" 

return val == missing or val == missingd.get(name) or val == '' 

 

def with_default_value(func, default): 

def newfunc(name, val): 

if ismissing(name, val): 

return default 

else: 

return func(val) 

return newfunc 

 

def mybool(x): 

if x == 'True': 

return True 

elif x == 'False': 

return False 

else: 

raise ValueError('invalid bool') 

 

dateparser = dateutil.parser.parse 

 

def mydateparser(x): 

# try and return a datetime object 

d = dateparser(x, dayfirst=dayfirst, yearfirst=yearfirst) 

return d 

 

mydateparser = with_default_value(mydateparser, datetime.datetime(1, 1, 1)) 

 

myfloat = with_default_value(float, np.nan) 

myint = with_default_value(int, -1) 

mystr = with_default_value(str, '') 

mybool = with_default_value(mybool, None) 

 

def mydate(x): 

# try and return a date object 

d = dateparser(x, dayfirst=dayfirst, yearfirst=yearfirst) 

 

if d.hour > 0 or d.minute > 0 or d.second > 0: 

raise ValueError('not a date') 

return d.date() 

mydate = with_default_value(mydate, datetime.date(1, 1, 1)) 

 

def get_func(name, item, func): 

# promote functions in this order 

funcs = [mybool, myint, myfloat, mydate, mydateparser, mystr] 

for func in funcs[funcs.index(func):]: 

try: 

func(name, item) 

except Exception: 

continue 

return func 

raise ValueError('Could not find a working conversion function') 

 

# map column names that clash with builtins -- TODO - extend this list 

itemd = { 

'return': 'return_', 

'file': 'file_', 

'print': 'print_', 

} 

 

def get_converters(reader, comments): 

 

converters = None 

i = 0 

for row in reader: 

if (len(row) and comments is not None and 

row[0].startswith(comments)): 

continue 

if i == 0: 

converters = [mybool]*len(row) 

if checkrows and i > checkrows: 

break 

i += 1 

 

for j, (name, item) in enumerate(zip(names, row)): 

func = converterd.get(j) 

if func is None: 

func = converterd.get(name) 

if func is None: 

func = converters[j] 

if len(item.strip()): 

func = get_func(name, item, func) 

else: 

# how should we handle custom converters and defaults? 

func = with_default_value(func, None) 

converters[j] = func 

return converters 

 

# Get header and remove invalid characters 

needheader = names is None 

 

if needheader: 

for row in reader: 

if (len(row) and comments is not None and 

row[0].startswith(comments)): 

continue 

headers = row 

break 

 

# remove these chars 

delete = set(r"""~!@#$%^&*()-=+~\|}[]{';: /?.>,<""") 

delete.add('"') 

 

names = [] 

seen = dict() 

for i, item in enumerate(headers): 

item = item.strip().lower().replace(' ', '_') 

item = ''.join([c for c in item if c not in delete]) 

if not len(item): 

item = 'column%d' % i 

 

item = itemd.get(item, item) 

cnt = seen.get(item, 0) 

if cnt > 0: 

names.append(item + '_%d' % cnt) 

else: 

names.append(item) 

seen[item] = cnt+1 

 

else: 

if isinstance(names, str): 

names = [n.strip() for n in names.split(',')] 

 

# get the converter functions by inspecting checkrows 

converters = get_converters(reader, comments) 

if converters is None: 

raise ValueError('Could not find any valid data in CSV file') 

 

# reset the reader and start over 

fh.seek(0) 

reader = csv.reader(fh, delimiter=delimiter) 

process_skiprows(reader) 

 

if needheader: 

while True: 

# skip past any comments and consume one line of column header 

row = next(reader) 

if (len(row) and comments is not None and 

row[0].startswith(comments)): 

continue 

break 

 

# iterate over the remaining rows and convert the data to date 

# objects, ints, or floats as appropriate 

rows = [] 

rowmasks = [] 

for i, row in enumerate(reader): 

if not len(row): 

continue 

if comments is not None and row[0].startswith(comments): 

continue 

# Ensure that the row returned always has the same nr of elements 

row.extend([''] * (len(converters) - len(row))) 

rows.append([func(name, val) 

for func, name, val in zip(converters, names, row)]) 

rowmasks.append([ismissing(name, val) 

for name, val in zip(names, row)]) 

fh.close() 

 

if not len(rows): 

return None 

 

if use_mrecords and np.any(rowmasks): 

r = np.ma.mrecords.fromrecords(rows, names=names, mask=rowmasks) 

else: 

r = np.rec.fromrecords(rows, names=names) 

return r 

 

 

# a series of classes for describing the format intentions of various rec views 

@cbook.deprecated("2.2") 

class FormatObj(object): 

def tostr(self, x): 

return self.toval(x) 

 

def toval(self, x): 

return str(x) 

 

def fromstr(self, s): 

return s 

 

def __hash__(self): 

""" 

override the hash function of any of the formatters, so that we don't 

create duplicate excel format styles 

""" 

return hash(self.__class__) 

 

 

@cbook.deprecated("2.2") 

class FormatString(FormatObj): 

def tostr(self, x): 

val = repr(x) 

return val[1:-1] 

 

 

@cbook.deprecated("2.2") 

class FormatFormatStr(FormatObj): 

def __init__(self, fmt): 

self.fmt = fmt 

 

def tostr(self, x): 

if x is None: 

return 'None' 

return self.fmt % self.toval(x) 

 

 

@cbook.deprecated("2.2") 

class FormatFloat(FormatFormatStr): 

def __init__(self, precision=4, scale=1.): 

FormatFormatStr.__init__(self, '%%1.%df' % precision) 

self.precision = precision 

self.scale = scale 

 

def __hash__(self): 

return hash((self.__class__, self.precision, self.scale)) 

 

def toval(self, x): 

if x is not None: 

x = x * self.scale 

return x 

 

def fromstr(self, s): 

return float(s)/self.scale 

 

 

@cbook.deprecated("2.2") 

class FormatInt(FormatObj): 

 

def tostr(self, x): 

return '%d' % int(x) 

 

def toval(self, x): 

return int(x) 

 

def fromstr(self, s): 

return int(s) 

 

 

@cbook.deprecated("2.2") 

class FormatBool(FormatObj): 

def toval(self, x): 

return str(x) 

 

def fromstr(self, s): 

return bool(s) 

 

 

@cbook.deprecated("2.2") 

class FormatPercent(FormatFloat): 

def __init__(self, precision=4): 

FormatFloat.__init__(self, precision, scale=100.) 

 

 

@cbook.deprecated("2.2") 

class FormatThousands(FormatFloat): 

def __init__(self, precision=4): 

FormatFloat.__init__(self, precision, scale=1e-3) 

 

 

@cbook.deprecated("2.2") 

class FormatMillions(FormatFloat): 

def __init__(self, precision=4): 

FormatFloat.__init__(self, precision, scale=1e-6) 

 

 

@cbook.deprecated("2.2", alternative='date.strftime') 

class FormatDate(FormatObj): 

def __init__(self, fmt): 

self.fmt = fmt 

 

def __hash__(self): 

return hash((self.__class__, self.fmt)) 

 

def toval(self, x): 

if x is None: 

return 'None' 

return x.strftime(self.fmt) 

 

def fromstr(self, x): 

import dateutil.parser 

return dateutil.parser.parse(x).date() 

 

 

@cbook.deprecated("2.2", alternative='datetime.strftime') 

class FormatDatetime(FormatDate): 

def __init__(self, fmt='%Y-%m-%d %H:%M:%S'): 

FormatDate.__init__(self, fmt) 

 

def fromstr(self, x): 

import dateutil.parser 

return dateutil.parser.parse(x) 

 

 

@cbook.deprecated("2.2") 

def get_formatd(r, formatd=None): 

'build a formatd guaranteed to have a key for every dtype name' 

defaultformatd = { 

np.bool_: FormatBool(), 

np.int16: FormatInt(), 

np.int32: FormatInt(), 

np.int64: FormatInt(), 

np.float32: FormatFloat(), 

np.float64: FormatFloat(), 

np.object_: FormatObj(), 

np.string_: FormatString()} 

 

if formatd is None: 

formatd = dict() 

 

for i, name in enumerate(r.dtype.names): 

dt = r.dtype[name] 

format = formatd.get(name) 

if format is None: 

format = defaultformatd.get(dt.type, FormatObj()) 

formatd[name] = format 

return formatd 

 

 

@cbook.deprecated("2.2") 

def csvformat_factory(format): 

format = copy.deepcopy(format) 

if isinstance(format, FormatFloat): 

format.scale = 1. # override scaling for storage 

format.fmt = '%r' 

return format 

 

 

@cbook.deprecated("2.2", alternative='numpy.recarray.tofile') 

def rec2txt(r, header=None, padding=3, precision=3, fields=None): 

""" 

Returns a textual representation of a record array. 

 

Parameters 

---------- 

r: numpy recarray 

 

header: list 

column headers 

 

padding: 

space between each column 

 

precision: number of decimal places to use for floats. 

Set to an integer to apply to all floats. Set to a 

list of integers to apply precision individually. 

Precision for non-floats is simply ignored. 

 

fields : list 

If not None, a list of field names to print. fields 

can be a list of strings like ['field1', 'field2'] or a single 

comma separated string like 'field1,field2' 

 

Examples 

-------- 

 

For ``precision=[0,2,3]``, the output is :: 

 

ID Price Return 

ABC 12.54 0.234 

XYZ 6.32 -0.076 

""" 

 

if fields is not None: 

r = rec_keep_fields(r, fields) 

 

if cbook.is_numlike(precision): 

precision = [precision]*len(r.dtype) 

 

def get_type(item, atype=int): 

tdict = {None: int, int: float, float: str} 

try: 

atype(str(item)) 

except: 

return get_type(item, tdict[atype]) 

return atype 

 

def get_justify(colname, column, precision): 

ntype = column.dtype 

 

if np.issubdtype(ntype, np.character): 

fixed_width = int(ntype.str[2:]) 

length = max(len(colname), fixed_width) 

return 0, length+padding, "%s" # left justify 

 

if np.issubdtype(ntype, np.integer): 

length = max(len(colname), 

np.max(list(map(len, list(map(str, column)))))) 

return 1, length+padding, "%d" # right justify 

 

if np.issubdtype(ntype, np.floating): 

fmt = "%." + str(precision) + "f" 

length = max( 

len(colname), 

np.max(list(map(len, list(map(lambda x: fmt % x, column))))) 

) 

return 1, length+padding, fmt # right justify 

 

return (0, 

max(len(colname), 

np.max(list(map(len, list(map(str, column))))))+padding, 

"%s") 

 

if header is None: 

header = r.dtype.names 

 

justify_pad_prec = [get_justify(header[i], r.__getitem__(colname), 

precision[i]) 

for i, colname in enumerate(r.dtype.names)] 

 

justify_pad_prec_spacer = [] 

for i in range(len(justify_pad_prec)): 

just, pad, prec = justify_pad_prec[i] 

if i == 0: 

justify_pad_prec_spacer.append((just, pad, prec, 0)) 

else: 

pjust, ppad, pprec = justify_pad_prec[i-1] 

if pjust == 0 and just == 1: 

justify_pad_prec_spacer.append((just, pad-padding, prec, 0)) 

elif pjust == 1 and just == 0: 

justify_pad_prec_spacer.append((just, pad, prec, padding)) 

else: 

justify_pad_prec_spacer.append((just, pad, prec, 0)) 

 

def format(item, just_pad_prec_spacer): 

just, pad, prec, spacer = just_pad_prec_spacer 

if just == 0: 

return spacer*' ' + str(item).ljust(pad) 

else: 

if get_type(item) == float: 

item = (prec % float(item)) 

elif get_type(item) == int: 

item = (prec % int(item)) 

 

return item.rjust(pad) 

 

textl = [] 

textl.append(''.join([format(colitem, justify_pad_prec_spacer[j]) 

for j, colitem in enumerate(header)])) 

for i, row in enumerate(r): 

textl.append(''.join([format(colitem, justify_pad_prec_spacer[j]) 

for j, colitem in enumerate(row)])) 

if i == 0: 

textl[0] = textl[0].rstrip() 

 

text = os.linesep.join(textl) 

return text 

 

 

@cbook.deprecated("2.2", alternative='numpy.recarray.tofile') 

def rec2csv(r, fname, delimiter=',', formatd=None, missing='', 

missingd=None, withheader=True): 

""" 

Save the data from numpy recarray *r* into a 

comma-/space-/tab-delimited file. The record array dtype names 

will be used for column headers. 

 

*fname*: can be a filename or a file handle. Support for gzipped 

files is automatic, if the filename ends in '.gz' 

 

*withheader*: if withheader is False, do not write the attribute 

names in the first row 

 

for formatd type FormatFloat, we override the precision to store 

full precision floats in the CSV file 

 

See Also 

-------- 

:func:`csv2rec` 

For information about *missing* and *missingd*, which can be used to 

fill in masked values into your CSV file. 

""" 

 

delimiter = str(delimiter) 

 

if missingd is None: 

missingd = dict() 

 

def with_mask(func): 

def newfunc(val, mask, mval): 

if mask: 

return mval 

else: 

return func(val) 

return newfunc 

 

if r.ndim != 1: 

raise ValueError('rec2csv only operates on 1 dimensional recarrays') 

 

formatd = get_formatd(r, formatd) 

funcs = [] 

for i, name in enumerate(r.dtype.names): 

funcs.append(with_mask(csvformat_factory(formatd[name]).tostr)) 

 

fh, opened = cbook.to_filehandle(fname, 'wb', return_opened=True) 

writer = csv.writer(fh, delimiter=delimiter) 

header = r.dtype.names 

if withheader: 

writer.writerow(header) 

 

# Our list of specials for missing values 

mvals = [] 

for name in header: 

mvals.append(missingd.get(name, missing)) 

 

ismasked = False 

if len(r): 

row = r[0] 

ismasked = hasattr(row, '_fieldmask') 

 

for row in r: 

if ismasked: 

row, rowmask = row.item(), row._fieldmask.item() 

else: 

rowmask = [False] * len(row) 

writer.writerow([func(val, mask, mval) for func, val, mask, mval 

in zip(funcs, row, rowmask, mvals)]) 

if opened: 

fh.close() 

 

 

@cbook.deprecated('2.2', alternative='scipy.interpolate.griddata') 

def griddata(x, y, z, xi, yi, interp='nn'): 

""" 

Interpolates from a nonuniformly spaced grid to some other grid. 

 

Fits a surface of the form z = f(`x`, `y`) to the data in the 

(usually) nonuniformly spaced vectors (`x`, `y`, `z`), then 

interpolates this surface at the points specified by 

(`xi`, `yi`) to produce `zi`. 

 

Parameters 

---------- 

x, y, z : 1d array_like 

Coordinates of grid points to interpolate from. 

xi, yi : 1d or 2d array_like 

Coordinates of grid points to interpolate to. 

interp : string key from {'nn', 'linear'} 

Interpolation algorithm, either 'nn' for natural neighbor, or 

'linear' for linear interpolation. 

 

Returns 

------- 

2d float array 

Array of values interpolated at (`xi`, `yi`) points. Array 

will be masked is any of (`xi`, `yi`) are outside the convex 

hull of (`x`, `y`). 

 

Notes 

----- 

If `interp` is 'nn' (the default), uses natural neighbor 

interpolation based on Delaunay triangulation. This option is 

only available if the mpl_toolkits.natgrid module is installed. 

This can be downloaded from https://github.com/matplotlib/natgrid. 

The (`xi`, `yi`) grid must be regular and monotonically increasing 

in this case. 

 

If `interp` is 'linear', linear interpolation is used via 

matplotlib.tri.LinearTriInterpolator. 

 

Instead of using `griddata`, more flexible functionality and other 

interpolation options are available using a 

matplotlib.tri.Triangulation and a matplotlib.tri.TriInterpolator. 

""" 

# Check input arguments. 

x = np.asanyarray(x, dtype=np.float64) 

y = np.asanyarray(y, dtype=np.float64) 

z = np.asanyarray(z, dtype=np.float64) 

if x.shape != y.shape or x.shape != z.shape or x.ndim != 1: 

raise ValueError("x, y and z must be equal-length 1-D arrays") 

 

xi = np.asanyarray(xi, dtype=np.float64) 

yi = np.asanyarray(yi, dtype=np.float64) 

if xi.ndim != yi.ndim: 

raise ValueError("xi and yi must be arrays with the same number of " 

"dimensions (1 or 2)") 

if xi.ndim == 2 and xi.shape != yi.shape: 

raise ValueError("if xi and yi are 2D arrays, they must have the same " 

"shape") 

if xi.ndim == 1: 

xi, yi = np.meshgrid(xi, yi) 

 

if interp == 'nn': 

use_nn_interpolation = True 

elif interp == 'linear': 

use_nn_interpolation = False 

else: 

raise ValueError("interp keyword must be one of 'linear' (for linear " 

"interpolation) or 'nn' (for natural neighbor " 

"interpolation). Default is 'nn'.") 

 

# Remove masked points. 

mask = np.ma.getmask(z) 

if mask is not np.ma.nomask: 

x = x.compress(~mask) 

y = y.compress(~mask) 

z = z.compressed() 

 

if use_nn_interpolation: 

try: 

from mpl_toolkits.natgrid import _natgrid 

except ImportError: 

raise RuntimeError( 

"To use interp='nn' (Natural Neighbor interpolation) in " 

"griddata, natgrid must be installed. Either install it " 

"from http://github.com/matplotlib/natgrid or use " 

"interp='linear' instead.") 

 

if xi.ndim == 2: 

# natgrid expects 1D xi and yi arrays. 

xi = xi[0, :] 

yi = yi[:, 0] 

 

# Override default natgrid internal parameters. 

_natgrid.seti(b'ext', 0) 

_natgrid.setr(b'nul', np.nan) 

 

if np.min(np.diff(xi)) < 0 or np.min(np.diff(yi)) < 0: 

raise ValueError("Output grid defined by xi,yi must be monotone " 

"increasing") 

 

# Allocate array for output (buffer will be overwritten by natgridd) 

zi = np.empty((yi.shape[0], xi.shape[0]), np.float64) 

 

# Natgrid requires each array to be contiguous rather than e.g. a view 

# that is a non-contiguous slice of another array. Use numpy.require 

# to deal with this, which will copy if necessary. 

x = np.require(x, requirements=['C']) 

y = np.require(y, requirements=['C']) 

z = np.require(z, requirements=['C']) 

xi = np.require(xi, requirements=['C']) 

yi = np.require(yi, requirements=['C']) 

_natgrid.natgridd(x, y, z, xi, yi, zi) 

 

# Mask points on grid outside convex hull of input data. 

if np.any(np.isnan(zi)): 

zi = np.ma.masked_where(np.isnan(zi), zi) 

return zi 

else: 

# Linear interpolation performed using a matplotlib.tri.Triangulation 

# and a matplotlib.tri.LinearTriInterpolator. 

from .tri import Triangulation, LinearTriInterpolator 

triang = Triangulation(x, y) 

interpolator = LinearTriInterpolator(triang, z) 

return interpolator(xi, yi) 

 

 

################################################## 

# Linear interpolation algorithms 

################################################## 

@cbook.deprecated("2.2", alternative="numpy.interp") 

def less_simple_linear_interpolation(x, y, xi, extrap=False): 

""" 

This function provides simple (but somewhat less so than 

:func:`cbook.simple_linear_interpolation`) linear interpolation. 

:func:`simple_linear_interpolation` will give a list of point 

between a start and an end, while this does true linear 

interpolation at an arbitrary set of points. 

 

This is very inefficient linear interpolation meant to be used 

only for a small number of points in relatively non-intensive use 

cases. For real linear interpolation, use scipy. 

""" 

x = np.asarray(x) 

y = np.asarray(y) 

xi = np.atleast_1d(xi) 

 

s = list(y.shape) 

s[0] = len(xi) 

yi = np.tile(np.nan, s) 

 

for ii, xx in enumerate(xi): 

bb = x == xx 

if np.any(bb): 

jj, = np.nonzero(bb) 

yi[ii] = y[jj[0]] 

elif xx < x[0]: 

if extrap: 

yi[ii] = y[0] 

elif xx > x[-1]: 

if extrap: 

yi[ii] = y[-1] 

else: 

jj, = np.nonzero(x < xx) 

jj = max(jj) 

 

yi[ii] = y[jj] + (xx-x[jj])/(x[jj+1]-x[jj]) * (y[jj+1]-y[jj]) 

 

return yi 

 

 

@cbook.deprecated("2.2") 

def slopes(x, y): 

""" 

:func:`slopes` calculates the slope *y*'(*x*) 

 

The slope is estimated using the slope obtained from that of a 

parabola through any three consecutive points. 

 

This method should be superior to that described in the appendix 

of A CONSISTENTLY WELL BEHAVED METHOD OF INTERPOLATION by Russel 

W. Stineman (Creative Computing July 1980) in at least one aspect: 

 

Circles for interpolation demand a known aspect ratio between 

*x*- and *y*-values. For many functions, however, the abscissa 

are given in different dimensions, so an aspect ratio is 

completely arbitrary. 

 

The parabola method gives very similar results to the circle 

method for most regular cases but behaves much better in special 

cases. 

 

Norbert Nemec, Institute of Theoretical Physics, University or 

Regensburg, April 2006 Norbert.Nemec at physik.uni-regensburg.de 

 

(inspired by a original implementation by Halldor Bjornsson, 

Icelandic Meteorological Office, March 2006 halldor at vedur.is) 

""" 

# Cast key variables as float. 

x = np.asarray(x, float) 

y = np.asarray(y, float) 

 

yp = np.zeros(y.shape, float) 

 

dx = x[1:] - x[:-1] 

dy = y[1:] - y[:-1] 

dydx = dy/dx 

yp[1:-1] = (dydx[:-1] * dx[1:] + dydx[1:] * dx[:-1])/(dx[1:] + dx[:-1]) 

yp[0] = 2.0 * dy[0]/dx[0] - yp[1] 

yp[-1] = 2.0 * dy[-1]/dx[-1] - yp[-2] 

return yp 

 

 

@cbook.deprecated("2.2") 

def stineman_interp(xi, x, y, yp=None): 

""" 

Given data vectors *x* and *y*, the slope vector *yp* and a new 

abscissa vector *xi*, the function :func:`stineman_interp` uses 

Stineman interpolation to calculate a vector *yi* corresponding to 

*xi*. 

 

Here's an example that generates a coarse sine curve, then 

interpolates over a finer abscissa:: 

 

x = linspace(0,2*pi,20); y = sin(x); yp = cos(x) 

xi = linspace(0,2*pi,40); 

yi = stineman_interp(xi,x,y,yp); 

plot(x,y,'o',xi,yi) 

 

The interpolation method is described in the article A 

CONSISTENTLY WELL BEHAVED METHOD OF INTERPOLATION by Russell 

W. Stineman. The article appeared in the July 1980 issue of 

Creative Computing with a note from the editor stating that while 

they were: 

 

not an academic journal but once in a while something serious 

and original comes in adding that this was 

"apparently a real solution" to a well known problem. 

 

For *yp* = *None*, the routine automatically determines the slopes 

using the :func:`slopes` routine. 

 

*x* is assumed to be sorted in increasing order. 

 

For values ``xi[j] < x[0]`` or ``xi[j] > x[-1]``, the routine 

tries an extrapolation. The relevance of the data obtained from 

this, of course, is questionable... 

 

Original implementation by Halldor Bjornsson, Icelandic 

Meteorolocial Office, March 2006 halldor at vedur.is 

 

Completely reworked and optimized for Python by Norbert Nemec, 

Institute of Theoretical Physics, University or Regensburg, April 

2006 Norbert.Nemec at physik.uni-regensburg.de 

""" 

 

# Cast key variables as float. 

x = np.asarray(x, float) 

y = np.asarray(y, float) 

if x.shape != y.shape: 

raise ValueError("'x' and 'y' must be of same shape") 

 

if yp is None: 

yp = slopes(x, y) 

else: 

yp = np.asarray(yp, float) 

 

xi = np.asarray(xi, float) 

yi = np.zeros(xi.shape, float) 

 

# calculate linear slopes 

dx = x[1:] - x[:-1] 

dy = y[1:] - y[:-1] 

s = dy/dx # note length of s is N-1 so last element is #N-2 

 

# find the segment each xi is in 

# this line actually is the key to the efficiency of this implementation 

idx = np.searchsorted(x[1:-1], xi) 

 

# now we have generally: x[idx[j]] <= xi[j] <= x[idx[j]+1] 

# except at the boundaries, where it may be that xi[j] < x[0] or 

# xi[j] > x[-1] 

 

# the y-values that would come out from a linear interpolation: 

sidx = s.take(idx) 

xidx = x.take(idx) 

yidx = y.take(idx) 

xidxp1 = x.take(idx+1) 

yo = yidx + sidx * (xi - xidx) 

 

# the difference that comes when using the slopes given in yp 

# using the yp slope of the left point 

dy1 = (yp.take(idx) - sidx) * (xi - xidx) 

# using the yp slope of the right point 

dy2 = (yp.take(idx+1)-sidx) * (xi - xidxp1) 

 

dy1dy2 = dy1*dy2 

# The following is optimized for Python. The solution actually 

# does more calculations than necessary but exploiting the power 

# of numpy, this is far more efficient than coding a loop by hand 

# in Python 

yi = yo + dy1dy2 * np.choose(np.array(np.sign(dy1dy2), np.int32)+1, 

((2*xi-xidx-xidxp1)/((dy1-dy2)*(xidxp1-xidx)), 

0.0, 

1/(dy1+dy2),)) 

return yi 

 

 

class GaussianKDE(object): 

""" 

Representation of a kernel-density estimate using Gaussian kernels. 

 

Parameters 

---------- 

dataset : array_like 

Datapoints to estimate from. In case of univariate data this is a 1-D 

array, otherwise a 2-D array with shape (# of dims, # of data). 

 

bw_method : str, scalar or callable, optional 

The method used to calculate the estimator bandwidth. This can be 

'scott', 'silverman', a scalar constant or a callable. If a 

scalar, this will be used directly as `kde.factor`. If a 

callable, it should take a `GaussianKDE` instance as only 

parameter and return a scalar. If None (default), 'scott' is used. 

 

Attributes 

---------- 

dataset : ndarray 

The dataset with which `gaussian_kde` was initialized. 

 

dim : int 

Number of dimensions. 

 

num_dp : int 

Number of datapoints. 

 

factor : float 

The bandwidth factor, obtained from `kde.covariance_factor`, with which 

the covariance matrix is multiplied. 

 

covariance : ndarray 

The covariance matrix of `dataset`, scaled by the calculated bandwidth 

(`kde.factor`). 

 

inv_cov : ndarray 

The inverse of `covariance`. 

 

Methods 

------- 

kde.evaluate(points) : ndarray 

Evaluate the estimated pdf on a provided set of points. 

 

kde(points) : ndarray 

Same as kde.evaluate(points) 

 

""" 

 

# This implementation with minor modification was too good to pass up. 

# from scipy: https://github.com/scipy/scipy/blob/master/scipy/stats/kde.py 

 

def __init__(self, dataset, bw_method=None): 

self.dataset = np.atleast_2d(dataset) 

if not np.array(self.dataset).size > 1: 

raise ValueError("`dataset` input should have multiple elements.") 

 

self.dim, self.num_dp = np.array(self.dataset).shape 

isString = isinstance(bw_method, str) 

 

if bw_method is None: 

pass 

elif (isString and bw_method == 'scott'): 

self.covariance_factor = self.scotts_factor 

elif (isString and bw_method == 'silverman'): 

self.covariance_factor = self.silverman_factor 

elif (np.isscalar(bw_method) and not isString): 

self._bw_method = 'use constant' 

self.covariance_factor = lambda: bw_method 

elif callable(bw_method): 

self._bw_method = bw_method 

self.covariance_factor = lambda: self._bw_method(self) 

else: 

raise ValueError("`bw_method` should be 'scott', 'silverman', a " 

"scalar or a callable") 

 

# Computes the covariance matrix for each Gaussian kernel using 

# covariance_factor(). 

 

self.factor = self.covariance_factor() 

# Cache covariance and inverse covariance of the data 

if not hasattr(self, '_data_inv_cov'): 

self.data_covariance = np.atleast_2d( 

np.cov( 

self.dataset, 

rowvar=1, 

bias=False)) 

self.data_inv_cov = np.linalg.inv(self.data_covariance) 

 

self.covariance = self.data_covariance * self.factor ** 2 

self.inv_cov = self.data_inv_cov / self.factor ** 2 

self.norm_factor = np.sqrt( 

np.linalg.det( 

2 * np.pi * self.covariance)) * self.num_dp 

 

def scotts_factor(self): 

return np.power(self.num_dp, -1. / (self.dim + 4)) 

 

def silverman_factor(self): 

return np.power( 

self.num_dp * (self.dim + 2.0) / 4.0, -1. / (self.dim + 4)) 

 

# Default method to calculate bandwidth, can be overwritten by subclass 

covariance_factor = scotts_factor 

 

def evaluate(self, points): 

"""Evaluate the estimated pdf on a set of points. 

 

Parameters 

---------- 

points : (# of dimensions, # of points)-array 

Alternatively, a (# of dimensions,) vector can be passed in and 

treated as a single point. 

 

Returns 

------- 

values : (# of points,)-array 

The values at each point. 

 

Raises 

------ 

ValueError : if the dimensionality of the input points is different 

than the dimensionality of the KDE. 

 

""" 

points = np.atleast_2d(points) 

 

dim, num_m = np.array(points).shape 

if dim != self.dim: 

raise ValueError("points have dimension {}, dataset has dimension " 

"{}".format(dim, self.dim)) 

 

result = np.zeros((num_m,), dtype=float) 

 

if num_m >= self.num_dp: 

# there are more points than data, so loop over data 

for i in range(self.num_dp): 

diff = self.dataset[:, i, np.newaxis] - points 

tdiff = np.dot(self.inv_cov, diff) 

energy = np.sum(diff * tdiff, axis=0) / 2.0 

result = result + np.exp(-energy) 

else: 

# loop over points 

for i in range(num_m): 

diff = self.dataset - points[:, i, np.newaxis] 

tdiff = np.dot(self.inv_cov, diff) 

energy = np.sum(diff * tdiff, axis=0) / 2.0 

result[i] = np.sum(np.exp(-energy), axis=0) 

 

result = result / self.norm_factor 

 

return result 

 

__call__ = evaluate 

 

 

################################################## 

# Code related to things in and around polygons 

################################################## 

@cbook.deprecated("2.2") 

def inside_poly(points, verts): 

""" 

*points* is a sequence of *x*, *y* points. 

*verts* is a sequence of *x*, *y* vertices of a polygon. 

 

Return value is a sequence of indices into points for the points 

that are inside the polygon. 

""" 

# Make a closed polygon path 

poly = Path(verts) 

 

# Check to see which points are contained within the Path 

return [idx for idx, p in enumerate(points) if poly.contains_point(p)] 

 

 

@cbook.deprecated("2.2") 

def poly_below(xmin, xs, ys): 

""" 

Given a sequence of *xs* and *ys*, return the vertices of a 

polygon that has a horizontal base at *xmin* and an upper bound at 

the *ys*. *xmin* is a scalar. 

 

Intended for use with :meth:`matplotlib.axes.Axes.fill`, e.g.,:: 

 

xv, yv = poly_below(0, x, y) 

ax.fill(xv, yv) 

""" 

if any(isinstance(var, np.ma.MaskedArray) for var in [xs, ys]): 

numpy = np.ma 

else: 

numpy = np 

 

xs = numpy.asarray(xs) 

ys = numpy.asarray(ys) 

Nx = len(xs) 

Ny = len(ys) 

if Nx != Ny: 

raise ValueError("'xs' and 'ys' must have the same length") 

x = xmin*numpy.ones(2*Nx) 

y = numpy.ones(2*Nx) 

x[:Nx] = xs 

y[:Nx] = ys 

y[Nx:] = ys[::-1] 

return x, y 

 

 

@cbook.deprecated("2.2") 

def poly_between(x, ylower, yupper): 

""" 

Given a sequence of *x*, *ylower* and *yupper*, return the polygon 

that fills the regions between them. *ylower* or *yupper* can be 

scalar or iterable. If they are iterable, they must be equal in 

length to *x*. 

 

Return value is *x*, *y* arrays for use with 

:meth:`matplotlib.axes.Axes.fill`. 

""" 

if any(isinstance(var, np.ma.MaskedArray) for var in [ylower, yupper, x]): 

numpy = np.ma 

else: 

numpy = np 

 

Nx = len(x) 

if not cbook.iterable(ylower): 

ylower = ylower*numpy.ones(Nx) 

 

if not cbook.iterable(yupper): 

yupper = yupper*numpy.ones(Nx) 

 

x = numpy.concatenate((x, x[::-1])) 

y = numpy.concatenate((yupper, ylower[::-1])) 

return x, y 

 

 

@cbook.deprecated('2.2') 

def is_closed_polygon(X): 

""" 

Tests whether first and last object in a sequence are the same. These are 

presumably coordinates on a polygonal curve, in which case this function 

tests if that curve is closed. 

""" 

return np.all(X[0] == X[-1]) 

 

 

@cbook.deprecated("2.2", message='Moved to matplotlib.cbook') 

def contiguous_regions(mask): 

""" 

return a list of (ind0, ind1) such that mask[ind0:ind1].all() is 

True and we cover all such regions 

""" 

return cbook.contiguous_regions(mask) 

 

 

@cbook.deprecated("2.2") 

def cross_from_below(x, threshold): 

""" 

return the indices into *x* where *x* crosses some threshold from 

below, e.g., the i's where:: 

 

x[i-1]<threshold and x[i]>=threshold 

 

Example code:: 

 

import matplotlib.pyplot as plt 

 

t = np.arange(0.0, 2.0, 0.1) 

s = np.sin(2*np.pi*t) 

 

fig, ax = plt.subplots() 

ax.plot(t, s, '-o') 

ax.axhline(0.5) 

ax.axhline(-0.5) 

 

ind = cross_from_below(s, 0.5) 

ax.vlines(t[ind], -1, 1) 

 

ind = cross_from_above(s, -0.5) 

ax.vlines(t[ind], -1, 1) 

 

plt.show() 

 

See Also 

-------- 

:func:`cross_from_above` and :func:`contiguous_regions` 

 

""" 

x = np.asarray(x) 

ind = np.nonzero((x[:-1] < threshold) & (x[1:] >= threshold))[0] 

if len(ind): 

return ind+1 

else: 

return ind 

 

 

@cbook.deprecated("2.2") 

def cross_from_above(x, threshold): 

""" 

return the indices into *x* where *x* crosses some threshold from 

below, e.g., the i's where:: 

 

x[i-1]>threshold and x[i]<=threshold 

 

See Also 

-------- 

:func:`cross_from_below` and :func:`contiguous_regions` 

 

""" 

x = np.asarray(x) 

ind = np.nonzero((x[:-1] >= threshold) & (x[1:] < threshold))[0] 

if len(ind): 

return ind+1 

else: 

return ind 

 

 

################################################## 

# Vector and path length geometry calculations 

################################################## 

@cbook.deprecated('2.2') 

def vector_lengths(X, P=2., axis=None): 

""" 

Finds the length of a set of vectors in *n* dimensions. This is 

like the :func:`numpy.norm` function for vectors, but has the ability to 

work over a particular axis of the supplied array or matrix. 

 

Computes ``(sum((x_i)^P))^(1/P)`` for each ``{x_i}`` being the 

elements of *X* along the given axis. If *axis* is *None*, 

compute over all elements of *X*. 

""" 

X = np.asarray(X) 

return (np.sum(X**(P), axis=axis))**(1./P) 

 

 

@cbook.deprecated('2.2') 

def distances_along_curve(X): 

""" 

Computes the distance between a set of successive points in *N* dimensions. 

 

Where *X* is an *M* x *N* array or matrix. The distances between 

successive rows is computed. Distance is the standard Euclidean 

distance. 

""" 

X = np.diff(X, axis=0) 

return vector_lengths(X, axis=1) 

 

 

@cbook.deprecated('2.2') 

def path_length(X): 

""" 

Computes the distance travelled along a polygonal curve in *N* dimensions. 

 

Where *X* is an *M* x *N* array or matrix. Returns an array of 

length *M* consisting of the distance along the curve at each point 

(i.e., the rows of *X*). 

""" 

X = distances_along_curve(X) 

return np.concatenate((np.zeros(1), np.cumsum(X))) 

 

 

@cbook.deprecated('2.2') 

def quad2cubic(q0x, q0y, q1x, q1y, q2x, q2y): 

""" 

Converts a quadratic Bezier curve to a cubic approximation. 

 

The inputs are the *x* and *y* coordinates of the three control 

points of a quadratic curve, and the output is a tuple of *x* and 

*y* coordinates of the four control points of the cubic curve. 

""" 

# TODO: Candidate for deprecation -- no longer used internally 

 

# c0x, c0y = q0x, q0y 

c1x, c1y = q0x + 2./3. * (q1x - q0x), q0y + 2./3. * (q1y - q0y) 

c2x, c2y = c1x + 1./3. * (q2x - q0x), c1y + 1./3. * (q2y - q0y) 

# c3x, c3y = q2x, q2y 

return q0x, q0y, c1x, c1y, c2x, c2y, q2x, q2y 

 

 

@cbook.deprecated("2.2") 

def offset_line(y, yerr): 

""" 

Offsets an array *y* by +/- an error and returns a tuple 

(y - err, y + err). 

 

The error term can be: 

 

* A scalar. In this case, the returned tuple is obvious. 

* A vector of the same length as *y*. The quantities y +/- err are computed 

component-wise. 

* A tuple of length 2. In this case, yerr[0] is the error below *y* and 

yerr[1] is error above *y*. For example:: 

 

import numpy as np 

import matplotlib.pyplot as plt 

 

x = np.linspace(0, 2*np.pi, num=100, endpoint=True) 

y = np.sin(x) 

y_minus, y_plus = mlab.offset_line(y, 0.1) 

plt.plot(x, y) 

plt.fill_between(x, y_minus, y2=y_plus) 

plt.show() 

 

""" 

if cbook.is_numlike(yerr) or (cbook.iterable(yerr) and 

len(yerr) == len(y)): 

ymin = y - yerr 

ymax = y + yerr 

elif len(yerr) == 2: 

ymin, ymax = y - yerr[0], y + yerr[1] 

else: 

raise ValueError("yerr must be scalar, 1xN or 2xN") 

return ymin, ymax