Upgrade tree-sitter to the latest version

[juli1]

What problem are you trying to solve?

We are using an old version of tree-sitter. We should upgrade to the latest.

What is your solution?

Upgrade tree-sitter.

Alternatives considered

Do not upgrade

What the reviewer should know

To upgrade Swift, we had to download the generated parser from their GitHub action and publish them on a repository. This is because the tree-sitter-swift maintainers do not want to publish the parser.c and scanner.c. We opened a bug report to ask them to do so but in the meantime, we have a workaround.

Testing

• large repo python: found one issue with python-code-style/function-naming. Fixed according to the new predicates reported here
• large repo JS/TS: more violations caught in our analyzer. After investigations, this is related to only one very long file. I suspect the issue is more related to parsing and bugfix than anything else.
• large repo Java: done, no difference
• large repo go: in no difference

Testing Methodology

1. Run the analyzer on previous and new version

datadog-static-analyzer  --directory path/to/repo --output result-[old|new].csv --format csv

If the same number of violations is found, all good. Otherwise, keep digging.

2. Sort the results

sort result-old.csv > results-old-sorted.csv

3. Check the difference

comm -3 results-old-sorted.csv results-new-sorted.csv

[alex-pinkus]

Rather than hosting the artifacts yourself , you might as well use the with-generated-files branch I already publish: https:/alex-pinkus/tree-sitter-swift/tree/with-generated-files

Of course, as I mentioned on the linked issue, you shouldn’t necessarily treat that as a long-term viable approach.

[juli1]

Rather than hosting the artifacts yourself , you might as well use the with-generated-files branch I already publish: https:/alex-pinkus/tree-sitter-swift/tree/with-generated-files

Of course, as I mentioned on the linked issue, you shouldn’t necessarily treat that as a long-term viable approach.

Thank you so much. I used the branch you provided, it's way cleaner this way. Thank you again!

[juli1]

I tested this new version against large repositories and there is an issue with the new tree-sitter update (e.g. this new version we try to use). One rule (python-code-style/function-naming) is not working correctly and it seems there is a regression in tree-sitter.

I opened a bug for tree-sitter here. I will keep this pull request open until we come to a conclusion with the tree-sitter maintainers (either we need to fix something or they need to fix something).