Add parameter to filter (include/exclude) files

[djbe]

Param to be able to filter files to parse in or out.

This would mitigate the problem about making ib scan XIBs and Storyboards for use cases when you only want storyboard scene constants and don't need to parse XIBs, but they would be parsed anyway and slightly impact performance on every build. Then storyboards would become an alias of ib --filter '*.storyboards'

The name of the argument is up for debate. --filter might be clearer than --exclude, because users might otherwise expect a --include parameter, which I don't know how we'd add.

[djbe]

I've looked into wath PathKit (and Foundation) supports, and it's not much. PathKit has a method for glob, but there's no globstar support (**), and the only thing I've found that supports it is:
https://github.com/Bouke/Glob
Problems with the latter are:

• this would be an extra dependency
• it doesn't seem very maintained
• the author wrote globstar support themselves, not with some system function/library

Instead, we might be better of only supporting regular expressions. If a user then wants to match (for example) all IB files in folders ending with "-iOS":
--filter ".*-iOS\.(storyboard|xib)"
It would be very easy even to allow for multiple filters, just check each path if it matches all regexes (internal or user provided).

[AliSoftware]

I'm ok with using RegExes instead of globs. Way easier to implement and wouldn't require another dependency.

Sure, it might not be as natural for users to write .*\.xcassets rather than just *.xcassets, that's true. But:

• using RegEx gives also more power in filtering
• once we support (re-)adding recursive folders search (passing a folder and letting SwiftGen search into it instead of giving exact path to file(s)), those will solve a good part of the use cases people would have for filters anyway
• That would be even more true if, like we stated in another issue, we start creating aliases for subcommands, like swiftgen xib == swiftgen ib --filter '.*\.xib', then users might have even less use cases where they'd need to provide their own custom filters anyway.

[djbe]

I did find this bit to convert glob patterns into regexes, the logic isn't that complicated (especially if we only support the "extended" part):
https://github.com/fitzgen/glob-to-regexp
But I'm not sure if it'd be worth the effort to support it (tests, maintain it, ...).

So my vote goes to regexes ✋.

[AliSoftware]

Nah, more risky and more to maintain for too little benefit imho†, so let's go with RegExes 👍

(† that's humble opinion for me, of course 😝 )

[djbe]

@AliSoftware seeing as #379 is ready to be merged (pending any remaining feedback), this will probably be the next item to be tackled. Not sure if it's a "requirement" for 6.0, we can always postpone this to 6.1.

---

Multiple filter types will exist:

• Some commands will have default filters, such as ib: .*\.(xib|nib|storyboard).
• Other commands should have an optional filter, such as yaml: .*\.ya?ml, but the user should be able to point to non-yaml extension files.
• The user should be able to filter out files with their own expressions: .*-iOS\.(storyboard|xib)

How will we combine these? Do user filters override the built-in ones, or do we combine (AND) them? I think overriding will be simpler and at the same time allow the user more flexibility.

We should write out all the default filters here (and maybe also the command aliases), so we don't need to discuss these in the PR.

[djbe]

Filters:

• colors: .*\.(clr|json|txt|xml)
• fonts: .*\.(otf|ttf|ttc)
• ib: .*\.(nib|storyboard|xib)
• storyboards: .*\.(storyboard)
• strings: .*\.(strings|stringsdict)
• xcassets: .*\.(xcassets)

Notes: currently we do the (font) matching based on the UTI type ("public.font"). I couldn't find any concrete list of supported extensions, there are some older ones such as dfont and other bizarre stuff. The user can always match those with a custom regex.

[AliSoftware]

1. Yes on user filters overriding the default ones. Especially imagine we forgot a compatible extension, AND-ing filters would not allow the user to add additional files to parse, only restrict them.

2. we don't support stringdict files (yet) so maybe we should postpone adding them in the default filter until we do?

3. For font files I remember forgetting some extensions back when the test was done on file extensions. That's why I switched to the UTI test iirc. Aren't there UTxxx functions available in the SDK to list known file extensions given an UTI though? Would allow us to cover them all?

4. aren't .nib files compiled XIBs? So they're not XML files we can parse are they?

[AliSoftware]

Once our Parser protocol supports filter and we define a default filter for each command, it would be nice and easy to print that default filter in the usage help btw

[AliSoftware]

PS : I want to release 6.0 ASAP. We know how we work and how little free time we have to work on SwiftGen between work and social life, is ready to have that "one last PR before the next release" end up staying in progress ("almost there!") For days then weeks and postpone too much.

I say we release early, release often. That's not really a breaking feature anyway, just an addition, so it can wait for 6.1 (even if 6.1 might come fast) so we don't postpone 6.0 anymore. There are a lot of features in master for which a lot of people have been waiting for too long already (especially the ready-to-merge JSON/YAML parser one)

[djbe]

2. Yeah, more of a reference for when it's implemented, I'll remove it.

3. Note that iOS only supports ttf/otf. I can't really find a "list" of extensions, except some articles such as https://www.macissues.com/2014/05/06/font-types-supported-in-os-x/, but even then some types are referred by name instead of extension. Apparently there are even "font suitcases" with no extension, so I guess we'll just have to make a decent "default", and the user can always override this.

4. Good point. Was just typing from memory about actual types in Xcode. Although... nibs can be loaded or maybe converted (back) to xib?

print that default filter in the usage help

Good idea

release soon

Totally agree. This shouldn't be a 6.0 blocker.