@iarna/toml

Better TOML parsing and stringifying all in that familiar JSON interface.

** TOML 1.0.0-rc.1 **

TOML Spec Support

The most recent version as of 2019-04-21: 1.0.0-rc.1

Other Versions

1.0.0-rc.1 parsers can load almost any TOML 0.4 and TOML 0.5 document, but
TOML 1.0.0-rc.1 docs are not always compatible with TOML 0.4 and TOML 0.5
parsers. If you're using this to generate TOML documents and you want an
older parser to be able to read them you may want to use the
latest TOML 0.5 version of this module.

Example

 const TOML = require('@iarna/toml')
const obj = TOML.parse(`[abc]
foo = 123
bar = [1,2,3]`)
/* obj =
{abc: {foo: 123, bar: [1,2,3]}}
*/
const str = TOML.stringify(obj)
/* str =
[abc]
foo = 123
bar = [ 1, 2, 3 ]
*/

Visit the project github for more examples!

Why @iarna/toml

• Support for TOML 1.0.0-rc.1!
• Highly correct! Careful adherence to spec.
• See TOML-SPEC-SUPPORT for a comparison of which TOML features
  are supported by the various Node.js TOML parsers.
• Speedy! See benchmarks at end.
• BigInt support on Node 10!
• 100% test coverage.
• Small parser bundle (if you use @iarna/toml/parse-string).
• No deps.
• Detailed and easy to read error messages‼

 > TOML.parse(src)
Error: Unexpected character, expecting string, number, datetime, boolean, inline array or inline table at row 6, col 5, pos 87:
5: "abc\"" = { abc=123,def="abc" }
6> foo=sdkfj
       ^
7:

TOML.parse(str) → Object (example)

Also available with: require('@iarna/toml/parse-string')

Synchronously parse a TOML string and return an object.

TOML.stringify(obj) → String (example)

Also available with: require('@iarna/toml/stringify)

Serialize an object as TOML.

[your-object].toJSON

If an object TOML.stringify is serializing has a toJSON method then it
will call it to transform the object before serializing it. This matches
the behavior of JSON.stringify.

The one exception to this is that toJSON is not called for Date objects
because JSON represents dates as strings and TOML can represent them natively.

moment objects are treated the
same as native Date objects, in this respect.

TOML.stringify.value(obj) -> String

Also available with: require('@iarna/toml/stringify').value

Serialize a value as TOML would. This is a fragment and not a complete
valid TOML document.

Promises and Streaming

The parser provides alternative async and streaming interfaces, for times
that you're working with really absurdly big TOML files and don't want to
tie-up the event loop while it parses.

TOML.parse.async(str[, opts]) → Promise(Object) (example)

Also available with: require('@iarna/toml/parse-async')

opts.blocksize is the amount text to parser per pass through the event loop. Defaults to 40kb.

Asynchronously parse a TOML string and return a promise of the resulting object.

TOML.parse.stream(readable) → Promise(Object) (example)

Also available with: require('@iarna/toml/parse-stream')

Given a readable stream, parse it as it feeds us data. Return a promise of the resulting object.

readable.pipe(TOML.parse.stream()) → Transform (example)

Also available with: require('@iarna/toml/parse-stream')

Returns a transform stream in object mode. When it completes, emit the
resulting object. Only one object will ever be emitted.

Lowlevel Interface (example) (example w/ parser debugging)

You construct a parser object, per TOML file you want to process:

 const TOMLParser = require('@iarna/toml/lib/toml-parser.js')
const parser = new TOMLParser()

Then you call the parse method for each chunk as you read them, or in a
single call:

 parser.parse(`hello = 'world'`)

And finally, you call the finish method to complete parsing and retrieve
the resulting object.

 const data = parser.finish()

Both the parse method and finish method will throw if they find a
problem with the string they were given. Error objects thrown from the
parser have pos, line and col attributes. TOML.parse adds a visual
summary of where in the source string there were issues using
parse-pretty-error and you can too:

 const prettyError = require('./parse-pretty-error.js')
const newErr = prettyError(err, sourceString)

What's Different

Version 3 of this module supports TOML 1.0.0-rc.1. Please see the
CHANGELOG for details on exactly whats changed.

TOML we can't do

• -nan is a valid TOML value and is converted into NaN. There is no way to
  produce -nan when stringifying. Stringification will produce positive nan.
• Detecting and erroring on invalid utf8 documents: This is because Node's
  UTF8 processing converts invalid sequences into the placeholder character
  and does not have facilities for reporting these as errors instead. We
  can detect the placeholder character, but it's valid to intentionally
  include them in documents, so erroring on them is not great.
• On versions of Node < 10, very large Integer values will lose precision.
  On Node >=10, bigints are used.
• Floating/local dates and times are still represented by JavaScript Date
  objects, which don't actually support these concepts. The objects
  returned have been modified so that you can determine what kind of thing
  they are (with isFloating, isDate, isTime properties) and that
  their ISO representation (via toISOString) are representative of their
  TOML value. They will correctly round trip if you pass them to
  TOML.stringify.
• Binary, hexadecimal and octal values are converted to ordinary integers and
  will be decimal if you stringify them.

Changes

I write a by hand, honest-to-god,
CHANGELOG
for this project. It's a description of what went into a release that you
the consumer of the module could care about, not a list of git commits, so
please check it out!

Benchmarks

You can run them yourself with:

 $ npm run benchmark

The results below are from my desktop using Node 13.13.0. The library
versions tested were @iarna/[email protected], [email protected], [email protected],
@sgarciac/[email protected], @ltd/[email protected], and [email protected].
The speed value is megabytes-per-second that the parser can process of that
document type. Bigger is better. The percentage after average results is
the margin of error.

New here is fast-toml. fast-toml is very fast, for some datatypes, but it
also is missing most error checking demanded by the spec. For 0.4, it is
complete except for detail of multiline strings caught by the compliance
tests. Its support for 0.5 is incomplete. Check out the
spec compliance doc
for details.

As this table is getting a little wide, with how npm and github display it,
you can also view it seperately in the
BENCHMARK document.

	@iarna/toml	toml-j0.4	toml	@sgarciac/bombadil	@ltd/j-toml	fast-toml
Overall	28MB/sec 0.55%	-	-	-	-	-
01-small-doc-mixed-type-inline-array	5.3MB/sec 0.48%	-	-	-	-	12MB/sec 0.13%
Spec Example: v0.4.0	25MB/sec 0.40%	9.9MB/sec 0.15%	0.9MB/sec 0.37%	1.3MB/sec 1.02%	28MB/sec 0.33%	-
Spec Example: Hard Unicode	63MB/sec 0.47%	17MB/sec 0.21%	2MB/sec 0.25%	0.6MB/sec 0.47%	65MB/sec 0.27%	79MB/sec 0.09%
Types: Array, Inline	7.2MB/sec 0.53%	4.1MB/sec 0.09%	0.1MB/sec 0.69%	1.4MB/sec 0.86%	10MB/sec 0.33%	9MB/sec 0.16%
Types: Array	6.8MB/sec 0.09%	6.8MB/sec 0.20%	0.2MB/sec 0.81%	1.3MB/sec 0.82%	8.9MB/sec 0.36%	29MB/sec 0.16%
Types: Boolean,	20MB/sec 0.22%	9.3MB/sec 0.29%	0.2MB/sec 0.91%	1.9MB/sec 0.85%	16MB/sec 0.29%	8.6MB/sec 0.22%
Types: Datetime	17MB/sec 0.09%	11MB/sec 0.17%	0.3MB/sec 0.75%	1.6MB/sec 0.42%	9.8MB/sec 0.40%	6.5MB/sec 0.11%
Types: Float	8.5MB/sec 0.29%	5.8MB/sec 0.33%	0.2MB/sec 0.91%	2.2MB/sec 0.91%	14MB/sec 0.25%	7.9MB/sec 0.33%
Types: Int	5.8MB/sec 0.13%	4.5MB/sec 0.14%	0.1MB/sec 0.63%	1.5MB/sec 0.73%	9.8MB/sec 0.14%	8.1MB/sec 0.16%
Types: Literal String, 7 char	25MB/sec 0.15%	8.3MB/sec 0.38%	0.2MB/sec 0.71%	2.3MB/sec 1.04%	23MB/sec 0.28%	14MB/sec 0.21%
Types: Literal String, 92 char	44MB/sec 0.23%	12MB/sec 0.14%	0.3MB/sec 0.63%	13MB/sec 1.12%	100MB/sec 0.14%	77MB/sec 0.15%
Types: Literal String, Multiline, 1079 char	23MB/sec 0.35%	7.2MB/sec 0.34%	0.9MB/sec 0.86%	47MB/sec 1.07%	380MB/sec 0.13%	641MB/sec 0.14%
Types: Basic String, 7 char	25MB/sec 0.09%	7MB/sec 0.08%	0.2MB/sec 0.82%	2.3MB/sec 1.02%	15MB/sec 0.12%	13MB/sec 0.14%
Types: Basic String, 92 char	44MB/sec 0.15%	8MB/sec 0.39%	0.1MB/sec 1.52%	12MB/sec 1.53%	70MB/sec 0.17%	71MB/sec 0.16%
Types: Basic String, 1079 char	24MB/sec 0.36%	5.7MB/sec 0.12%	0.1MB/sec 3.65%	42MB/sec 1.67%	93MB/sec 0.13%	617MB/sec 0.14%
Types: Table, Inline	9.4MB/sec 0.21%	5.2MB/sec 0.23%	0.1MB/sec 1.18%	1.4MB/sec 1.20%	8.5MB/sec 0.68%	8.7MB/sec 0.30%
Types: Table	6.8MB/sec 0.13%	5.5MB/sec 0.22%	0.1MB/sec 1.10%	1.5MB/sec 1.05%	7.3MB/sec 0.54%	19MB/sec 0.21%
Scaling: Array, Inline, 1000 elements	40MB/sec 0.27%	2.4MB/sec 0.20%	0.1MB/sec 1.90%	1.6MB/sec 1.14%	18MB/sec 0.16%	32MB/sec 0.12%
Scaling: Array, Nested, 1000 deep	2MB/sec 0.17%	1.6MB/sec 0.09%	0.3MB/sec 0.62%	-	1.8MB/sec 0.80%	13MB/sec 0.19%
Scaling: Literal String, 40kb	59MB/sec 0.26%	10MB/sec 0.14%	3MB/sec 0.91%	13MB/sec 0.40%	479MB/sec 0.25%	19kMB/sec 0.20%
Scaling: Literal String, Multiline, 40kb	61MB/sec 0.23%	5.3MB/sec 0.30%	0.2MB/sec 1.78%	12MB/sec 0.55%	276MB/sec 0.16%	21kMB/sec 0.10%
Scaling: Basic String, Multiline, 40kb	61MB/sec 0.21%	6MB/sec 0.40%	2.8MB/sec 0.75%	12MB/sec 0.60%	1kMB/sec 0.13%	27kMB/sec 0.14%
Scaling: Basic String, 40kb	60MB/sec 0.13%	6.6MB/sec 0.13%	0.2MB/sec 1.67%	13MB/sec 0.30%	504MB/sec 0.26%	19kMB/sec 0.22%
Scaling: Table, Inline, 1000 elements	26MB/sec 0.17%	7.3MB/sec 0.83%	0.3MB/sec 0.95%	2.5MB/sec 1.24%	5.4MB/sec 0.22%	13MB/sec 0.22%
Scaling: Table, Inline, Nested, 1000 deep	8MB/sec 0.10%	5.2MB/sec 0.25%	0.1MB/sec 0.45%	-	3.1MB/sec 0.58%	10MB/sec 0.19%

Tests

The test suite is maintained at 100% coverage:

The spec was carefully hand converted into a series of test framework
independent (and mostly language independent) assertions, as pairs of TOML
and YAML files. You can find those files here:
spec-test.

Further tests were written to increase coverage to 100%, these may be more
implementation specific, but they can be found in coverage and
coverage-error.

I've also written some quality assurance style tests, which don't contribute
to coverage but do cover scenarios that could easily be problematic for some
implementations can be found in:
test/qa.js and
test/qa-error.js.

All of the official example files from the TOML spec are run through this
parser and compared to the official YAML files when available. These files are from the TOML spec as of:
357a4ba6
and specifically are:

• github.com/toml-lang/toml/tree/357a4ba6/examples
• github.com/toml-lang/toml/tree/357a4ba6/tests

The stringifier is tested by round-tripping these same files, asserting that
TOML.parse(sourcefile) deepEqual
TOML.parse(TOML.stringify(TOML.parse(sourcefile)). This is done in
test/roundtrip-examples.js
There are also some tests written to complete coverage from stringification in:
test/stringify.js

Tests for the async and streaming interfaces are in test/async.js and test/stream.js respectively.

Tests for the parser's debugging mode live in test/devel.js.

And finally, many more stringification tests were borrowed from @othiym23's
toml-stream module. They were fetched as of
b6f1e26b572d49742d49fa6a6d11524d003441fa and live in
test/toml-stream.

Improvements to make

• In stringify:
  • Any way to produce comments. As a JSON stand-in I'm not too worried
    about this. That said, a document orientated fork is something I'd like
    to look at eventually…
  • Stringification could use some work on its error reporting. It reports
    what's wrong, but not where in your data structure it was.
• Further optimize the parser:
  • There are some debugging assertions left in the main parser, these should be moved to a subclass.
  • Make the whole debugging parser thing work as a mixin instead of as a superclass.