From JavaScript or TypeScript

npm install @appliedtopology/tda4j

import { persistence } from "@appliedtopology/tda4j";

const points = [[0, 0], [1, 0], [1, 1], [0, 1]];
const result = persistence(points, { maxDimension: 1 });

result.bars;                // [{ dim: 0, birth: 0, death: Infinity }, { dim: 1, birth: 1, death: 1.41... }, ...]
result.representative(1);   // { cells: [[0, 1], [1, 2], ...], coefficients: [...] }: the loop of the bar result.bars[1]
result.plotBarcode().svg(); // an SVG document, to put in a page or a file

The functions take arrays of numbers and an options object, and a bar is { dim, birth, death } with death equal to Infinity for a class that never dies. They are the MATLAB and Java interface in JavaScript clothes, and the options are its options, spelled the same way: complex ("vr" is the default; "cech", "alpha", "witness", "dtm-rips", "dtm-alpha" and "sparse-rips" too), engine, maxDimension (default 2), maxFiltrationValue, prime (default 17), field, representativeType, minPersistence, includeZeroLength and the rest, which index.d.ts in the package lists with their types. A bar much shorter than the scale of the data is hidden, as in the other interfaces: result.allBars has every bar, and result.hiddenCount says how many were left out.

A Result also has bottleneckDistance(other, dimension), wassersteinDistance(other, dimension, order), landscape(dimension, numLevels, tMin, tMax, resolution) and the plots plotBarcode(), plotDiagram(), plotComplex(at) and plotRepresentative(i, at). A plot has svg() and html() for the text of a picture or a page, show() to put it in the page (in a browser) or open it in the default browser (from Node.js), and save(path).

A Result also has persistenceImage(dimension, sigma, birthMin, birthMax, persistenceMin, persistenceMax, birthResolution, persistenceResolution), and the complex it was computed from: numCells, boundary() (the nonzero entries as { rows, cols, values }, cells in filtration order) and, for cell j, columnDimension(j), columnVertices(j) and columnFiltrationValue(j).

Landmarks, circular coordinates and the text of files have functions of their own:

The parsers take the text, so a page can read a file the user picked with file.text().

A bad option or a malformed input throws an Error that says what to change.

On the command line

npx @appliedtopology/tda4j --max-dimension 1 points.csv runs the command line of the JVM build, with the same flags, reading files through Node.js; once the package is installed, tda4j --max-dimension 1 points.csv does. It does not read image files (--input-format image): decode the image yourself and pass its pixels to persistenceFromImage.

What differs from the JVM build

• One thread. JavaScript has no threads to share the work between. The alpha triangulation, the Cech and related filtration values and the other steps the JVM build shares between threads give the same answer, not faster; to keep a page responsive, run the computation in a Web Worker.
• Files need Node.js. In a browser there is no file system: save, and the readers of the file formats, throw an error that says so. Pass the data in memory.
• The live viewer is the page. plot.show() appends the plot to the page in a browser, and opens it in the default browser from Node.js; there is no viewer server.
• Speed. The engines run at their usual pace on a few thousand simplices; programs that work with hundreds of millions of simplices belong on the JVM, where the engines were tuned and memory is not limited by the browser.
• Printed numbers. A number prints the JavaScript way: 1.0 as 1, 1.0E-7 as 1e-7, where the JVM prints 1.0 and 1.0E-7. The command line, Show, the file writers and saved results are affected; every text reads back to the same number on both platforms, but a file written on one is not byte-identical to the other's.
• Floating point. Sums and square roots are the same IEEE operations as on the JVM, but Math.pow and friends are the JavaScript engine's own, so a value computed through them can differ in the last digit.

From Scala.js

In a Scala.js project, the dependency is the same line as for the JVM (see installation); sbt 2's %% picks the Scala.js artifact tda4j_sjs1_3 by itself (%%% in sbt 1). The code is the Scala code of the quickstart: Persistence(points) works as it does on the JVM. The cross-build's layout is described in the Developer's Guide architecture page.