Patrick Thomson
Patrick Thomson
_Ported from https://github.com/tree-sitter/haskell-tree-sitter/issues/99_ In the a-la-carte AST world, we dealt with comments by reifying them as nodes in the AST; since ALC ASTs have a degree of heterogeneity w/r/t typing,...
Now that we’re on GH Actions, we can pull in a licensed executable no sweat. We should do that, as we used to before the OSS days.
Assuming `a` is a record and `c` is an already-bound value, this reasonable-looking code: ``` a.b = c ``` fails under concrete evaluation, because it attempts to lookup `a.b` first,...
The FileCheck-esque testing framework for `semantic-python` fixtures has proven versatile. However, it doesn’t offer a pretty fundamental characteristic: the ability to run a compiled program and verify its output. This...
A conversation in a Slack: > robrix: the problem is that the scope graph analysis evals the body eagerly > robrix: but the variable isn’t initialized until analyzing the function...
This function, as @robrix pointed out, is a little weird: its second parameter can be a file path rather than a directory, in which case the resulting project’s root is...
For the semantic-python test suite, we’ve instituted a FileCheck-esque syntax, allowing us to put `jq(1)` statements that let us make assertions about the shape of some translated Python code and...
The use of `Error SomeException` is, in my estimation, an antipattern. It makes it too difficult to have fine-grained, principled extension boundaries. We should only use `Error` on recoverable exceptions,...
We should pull in something like https://github.com/nomeata/gipeda and graph our performance over time.
This should be a matter of augmenting the options definition.