.env ↔ JSON

Both, chosen by the mode option, which defaults to env-to-json. That direction reads a .env file and gives a flat JSON object printed with 2-space indentation. The json-to-env direction takes a flat JSON object and writes KEY=value lines, one per key, in the order the object lists them.

Lines of the form KEY=VALUE, where KEY starts with a letter or underscore and continues with letters, digits or underscores. An optional export prefix is accepted. Blank lines and lines starting with # are skipped, and any other line that does not match is skipped silently rather than causing an error.

A matching pair of single or double quotes wrapping the whole value is removed, and everything inside is taken as written. There is no escape processing, so a backslash-n inside the quotes stays two characters, and a multi-line value written across several lines will not be read as one value.

No. Values are copied as written, so a reference to another variable stays as literal text in the JSON. Nothing is resolved, substituted or read from your actual environment, and no shell is involved. That also means a value which depends on another one will not be expanded on the way back either.

It quotes a value only when it contains whitespace, a double quote or a hash, and escapes any inner double quotes. Values that are not strings are written through JSON stringify, so numbers and booleans become their plain text and an object becomes JSON on one line. A top-level array or non-object is refused with "expected a flat JSON object of key/value pairs".

The keys and values survive; the presentation does not. Comments, blank lines, the export prefix and the original ordering of quoting are not preserved, and if a key appears twice the last one wins. Also, your .env likely holds secrets, which is a good reason to note that this runs entirely in a Web Worker on your device.