NumpyArray: Binary Numpy data in Javascript

Description

This package provides a simple interface for loading binary Numpy array data (.npy) from URL and reading them in an elegant fashion. The Numpy data can be single-typed or mixed-typed (aka. structured).

Numpy arrays data (.npy) is usually generated by the Python scientific computing package Numpy.

Demo

https://gywn.github.io/NumpyArray.js/

Usage

→ NumpyArray.getFromURL(url[, header])

Load Numpy arrays data from url and return a Promise object. Once loaded, the array is given as the first argument for the fulfillment handler.

An optional header can be given to describe the structure of the Numpy arrays data. Otherwise the package will try to use the header object extracted from the binary data.

getFromURL only supports GET method.

Example

const promise = NumpyArray.getFromURL('npy/World Population Density.npy');
promise.then(na => window['na'] = na);

→ new NumpyArray(buf[, header])

Construct a NumpyArray object directly from an ArrayBuffer object.

Example

const r = new XMLHttpRequest();
r.open('GET', 'npy/World Population Density.npy');
r.responseType = 'arraybuffer';
r.onload = () => window['na'] = new NumpyArray(r.response);
r.send();

→ .fieldNames

Return an array of the fields' name. If the array is single-typed, the fieldNames is always ["value"], unless overriden by the header argument.

Example

na.fieldNames;  // ["Country", "Population", "Density"]

→ .size

Return the number of records. If the number of dimensions of the data is higher than 1, it will be flatten according to header.fortran_order.

Example

na.size;  // 197

→ .get(...)

Extract columns or rows from the data. The returned value is also a NumpyArray object (which is a view on the buffer).

Example

// Extract columns
na.get('Country').fieldNames  // ["Country"]
na.get(['Density', 'Country']).size  // 197

// Extract rows
na.get(99)  // ["Liechtenstein", 35236, 219.430419921875]
na.get([195, 196])  // [["Zambia", 13881336, 18.444271087646484],
                    //  ["Zimbabwe", 12084304, 30.93951988220215]]

→ .toArray()

Return a copy the data as Javascript Array object.

Example

na.toArray()  // [["Afghanistan", 29835392, 46.0780029296875], ...]
na.get('Country').toArray()  // ["Afghanistan", "Albania", ...]

→ .numpyHeader

Return the raw bytes of the header from the data. Could be useful if the package failed to extract the header object automatically and thus a header object must be given explicitly.

Example

na.numpyHeader  // "{'descr': [('Country', '|S22'),
                //             ('Population', '<i4'),
                //             ('Density', '<f4')],
                //   'fortran_order': False,
                //   'shape': (197,),
                //  }"