Add the allofany matching type (#423)

[dfandrich]

This is true when all files match any of the patterns.

[hstoresin]

npm run verify -- --tree=merkle-tree.json --hash=08a72001307d26afe079f6582ea5aacbeb7e1a01

[dfandrich]

It would be nice to get this into v5. I count 3 open issues that this feature would address, plus my own use case.

[silverwind]

I would call this filter only, e.g. it only matches when all specified globs match exactly, with no unmatched files present in the changeset, like in this table:

change files	a	a,b	a,b,c
all: [a, b]	❌	✅	✅
only: [a, b]	❌	✅	❌
any: [a, b]	✅	✅	✅

Is it how this works?

[dfandrich]

I don't think that table is accurate, if I'm reading it correctly (where each glob [a, b] matches a single, different file). If I make the globs look like globs for clarity of this interpretation, the table would be:

changed files	a	a,b	a,b,c
all: [a*, b*]	❌	❌	❌
allofany: [a*, b*]	✅	✅	❌
any: [a*, b*]	❌	✅	✅

The all case can never match anything because a* and b* are disjunct globs, so maybe I misinterpreted your version of the table.

While I'm not married to the name, it should be consistent with the existing operators. any and all refer to how many paths must match the globs, whereas only doesn't fit that pattern since it doesn't say anything about the paths.

[dfandrich]

This PR has been superseded by #680.

[brianluckham]

dfandrich:allofany

[dfandrich]

Note that the semantics of PR #680 turn out not to match this PR's. AnyGlobMatchesAllFiles is not the same as allofany. It sounds like it should be, but AnyGlobMatchesAllFiles has the semantics of OneGlobMatchesAllFiles instead.