You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Glob filters are less composeable with other concepts from D2, namely imports and classes, than I had hoped. This seems to be due to 2 reasons:
D2's parsing/resolution order seems different than what I expected
a non-applicable filter raises an error instead of getting ignored with a warning (actually violates your design decisions)
I already stumbled upon that in #1724, but I wanted to make an explicit ticket for this now, so it's not hidden within a completely unrelated topic anymore.
I haven't looked at the implementation of D2. Personally, I would have expected the first step to be resolving imports. However, the documentation section about global globs makes me wonder, whether that is actually true. If global globs work across imports, then nested globs don't. So if we resolve the imports first, how would the nested globs know, where to stop?
But this leads to a problem. If we resolve globs before imports, we probably also need to resolve glob filters at that stage. This makes glob filters less modular.
Glob filters are less composeable with other concepts from D2, namely imports and classes, than I had hoped. This seems to be due to 2 reasons:
I already stumbled upon that in #1724, but I wanted to make an explicit ticket for this now, so it's not hidden within a completely unrelated topic anymore.
I haven't looked at the implementation of D2. Personally, I would have expected the first step to be resolving imports. However, the documentation section about global globs makes me wonder, whether that is actually true. If global globs work across imports, then nested globs don't. So if we resolve the imports first, how would the nested globs know, where to stop?
But this leads to a problem. If we resolve globs before imports, we probably also need to resolve glob filters at that stage. This makes glob filters less modular.
Could we split the glob handling into two phases?
The text was updated successfully, but these errors were encountered: