untruncate-json

Fix up the end of a partial JSON string to create valid JSON.


Motivation

Have you ever been given a string that started off as valid JSON, but was then
truncated to fit under some maximum length? And even though it was invalid JSON
because it was cut off in the middle, you still wanted to parse it as JSON
anyways so you could extract what information you could from it?

No?

Well if you ever do, then this is the library for you!

Installation

With NPM:

``
npm install untruncate-json
`

With Yarn:

`
yarn add untruncate-json
`

Usage

Import the untruncateJson function:

`ts
import untruncateJson from "untruncate-json";
`

Call it on a truncated JSON string to get a complete, valid JSON string:

`ts
untruncateJson("[1, 2"); // -> "[1, 2]"
untruncateJson('"Hello, Wor'); // -> '"Hello, Wor"'
untruncateJson('{"votes": [true, fa'); // -> '{"votes": [true, false]}'
untruncateJson("123."); // -> "123.0"
`

untruncateJson will sometimes remove characters as well, if there was no way
to infer what value they were starting:

`ts
untruncateJson("[1, 2, "); // -> "[1, 2]"
untruncateJson('"abc\\'); // -> '"abc"'
untruncateJson('{"x": 20, "y": '); // -> '{"x": 20}'
`

Check the test
cases
for many more examples.

Guarantees

- As long as the input is the prefix of a valid JSON string and is not entirely
whitespace, the output will be valid JSON.
- This takes into account all parts of the JSON spec, including rules around
exponents in numbers, Unicode escape sequences in strings, and so forth.
- untruncateJson does not attempt to validate the input JSON. Therefore if
the input JSON is not the prefix of valid JSON, then the output will not be
valid JSON and will fail if you try to parse it.
- The output will always contain as many correct values as can be determined
under the assumption that the input was indeed the prefix of valid JSON. For
example, if the input is cut off after a bare t, then truncateJson will
assume that it was the start of the token true`.
- The last value appearing in the output JSON may have an incorrect value, for
example if it is a truncated number or string. However, it will never have the
incorrect _type_, e.g. if the input contained a string in that position, so
will the output.
- In cases where the type of the last value cannot be inferred, then characters
are removed from the end of the input as needed to avoid making a potentially
incorrect guess. For example, if the truncation occurs after an object key but
before any part of its corresponding value, then the entire key will be
removed.

Copyright © 2019 David Philipson