vweevers / tabular-stream

Detects tabular data (spreadsheets, dsv or json, 20+ different formats) and emits normalized objects.

tabular-stream

Note: the core of tabular-stream 1.0 has moved to detect-tabular. This is now a data normalizer on top of that.

Detects tabular data (spreadsheets, dsv or json, 20+ different formats) and emits objects. Ensures all rows have the same keys, optionally transforms keys and tries to coerce values to numbers. Spreadsheets and DSV must have a header.

Example

npm i tabular-stream snake-case format-data

const tabular = require('tabular-stream')
const fs = require('fs')
const snake = require('snake-case').snakeCase
const format = require('format-data')

fs.createReadStream('test/air_pollution_nl.xlsx')
  .pipe(tabular(snake))
  .pipe(format('json'))
  .pipe(process.stdout)

Need a CLI doing just this? Jump to tabular-cli, which pairs tabular-stream with format-data to convert tabular data to json, ndjson, dsv or sse. For example:

tabular -k snake-case -o tsv < input.xls > output.tsv

API

tabular([keys || options])

Returns a duplex stream - give it any tabular data, get back objects. (keys) is a shorthand for ({ keys: keys }). The available options are:

function keys

An optional function to transform and/or filter keys. Receives a single argument, for every key of the first row. Everything at change-case works well. If it returns an empty string or anything other than a string, the key is ignored (i.e. not included in the emitted objects).

function keys(key) {
  if (key === 'useless') return false
  return key.toUpperCase()
}

mixed defaultValue

Fallback value to use for null and undefined values. Default is 0.

boolean bare

Whether to emit null prototype objects via Object.create(null) or plain javascript objects (the default).

Other

Other options are passed as-is to spreadsheet-stream (if applicable). NB. Because the binary spreadsheets formats are not streamable, spreadsheet-stream will buffer the whole thing in memory. As a safe-guard you can set the maxSize option (in bytes): tabular({ maxSize: 1024 * 1024 }). See spreadsheet-stream for details.

Install

With npm do:

npm install tabular-stream

License

MIT.

Test data © Statistics Netherlands, The Hague/Heerlen.

tabular-stream's Issues

Tests fail with excel-stream (incorrect output)

Three problems:

• Object keys weirdly include a " character (courtesy of csv-stream, not an issue if csv-parser is used)
• excel-stream coerces to numbers..
  • but does not account for european notation (1,23 instead of 1.23)
  • and my test does not expect numbers
• Output of j ends with two newlines, and both csv-stream and csv-parser turn that into an object with empty values

Data omitted if not in first row of XLS

Using an XLS file (not a CSV), if columns are missing from the first data row (second row) then you only get the data on subsequent rows for the columns present in the first data row.

Consider the following, but saved as an XLS file:

A,B,C
1,,
1,2,3
4,5,6

You get:

{ A: 1 }
{ A: 1 }
{ A: 4 }

Using detect-tabular instead you get:

{ A: 1 }
{ A: 1, B: 2, C: 3 }
{ A: 4, B: 5, C: 6 }

Reading a CSV file gives you:

{ A: 1, B: 0, C: 0 }
{ A: 1, B: 2, C: 3 }
{ A: 4, B: 5, C: 6 }

Ignore the issue with the zeros.

It may be that the underlying Excel reader could be considered faulty and should emit consistent objects with null values for missing data, but I'm not sufficiently familiar with it all to make a judgement.

Test code that I used:

var tabular = require("tabular-stream"); // or "detect-tabular" for comparison.
var fs = require("fs");
var stream = tabular();
stream.on("data", console.log);
stream.end(fs.readFileSync(__dirname+"/test.xls"));