Tests of various grammars fail in locales with "," decimal separator

[th-we]

This issue occurs in a number of grammars, so I'm raising it here first before creating pull requests for the individual repos.

When running npm test e.g. on the tree-sitter-python grammar, the test fails:

script/parse-examples: Zeile 38: printf: 100.00000000000000000000: Ungültige Zahl.
Successfully parsed 0,00% of examples/numpy (407 of 407 files)
npm ERR! Test failed.  See above for more details.

This is my current setup:

$ echo "$OSTYPE"
linux-gnu
$ echo "$LANG"
de_DE.UTF-8
$ echo "$BASH_VERSION"
4.4.20(1)-release

The issue is caused by the script/parse-examples scripts of the Haskell, Ruby, Python, JavaScript, Rust, Go, Java and Julia grammars.

Options for fixing this:

1. set LC_NUMERIC=C in the script
2. in package.json, change the test script call to tree-sitter test && LANG=C script/parse-examples
3. at least document that tests should be run like LANG=C npm test

If there is an agreement, I'll create pull requests. Option 1 seems the most robust solution to me.

[maxbrunsfeld]

I'd be ok with adding LC_NUMERIC=C. I'd also be fine with just simplifying down to integer percentages, and avoiding floating point numbers entirely.

[maxbrunsfeld]

I'm going to close this out, since it's not an issue with this repo. If you're interested in opening PRs to fix the shell scripts, I'd merge them.