zip: Entry.name(); Entry.extract_to() accepting user-provided writers

[l1yefeng]

std.zip users sometimes wish to extract content in memory rather than to the file system. This change adds a new extract_to as a more flexible alternative to the existing extract.

This partly addresses #21922.

I am a first-time contributor to zig. Comments and suggestions are greatly appreciated.

[GalaxyShard]

Disclaimer: I'm only a contributor, not a Zig core team member, so take my suggestions with a grain of salt.

[GalaxyShard]

The purpose of this openDir function really isn't to open a directory, but to create directories to form that path, which is evident by the filesystem implementation of openDir below and the original code that was here before these changes.

There's also not much reason to have a close function; any extractor implementation could just call it before returning from it's openDir (or more accurately makePath) function if it was necessary. The filesystem implementation implements it as a no-op anyways.

[l1yefeng]

I agree with this.

[GalaxyShard]

Just a thought, but I do wonder if this could be factored out to remove the need for this extractor interface altogether, since all that's really needed here is a writer to decompress the data to.

For example, maybe have extract_to return a struct containing the filename/path, crc, and a "file or directory" enum, then this code that actually does the decompression and the openDir code above that creates the directory could be moved into extract, and the extractor interface can be removed.

[l1yefeng]

Decompression depends on the (generalized concept of) opened file and opening the file depends on the filename. Then decompression may have to be moved from extract_to to its caller, too. That's too much for callers I'm afraid.

[GalaxyShard]

Since this function creates a file and ensures it's parent directory exists, and doesn't open a file in the usual sense, it should probably be renamed as such. Of course if the extractor interface is removed (see comment above), this wouldn't be relevant.

[l1yefeng]

After using zip.zig a bit more, I realize that what makes things easier is to have these methods

• entry.name() that gets the filename. Often callers need the name first. Whether to extract and how to extract depends on the name. On the other hand, callers may also choose to extract_to straight away if they don't care about the name, being rid of the filename buffer.
• entry.extract_to() that accepts a simple writer instead of extract_context above. Callers can init/deinit as they wish at call sites.
• entry.extract() still wraps extract_to, roughly goes like

name <- self.name()
if (name ends with '/') {
  create dir
} else {
  file <- create file with name
  defer file.close()
  self.extract_to(..., file.writer())
}