use case: ability to declare dependency on single file

[Techatrix]

ZLS depends on Zig's langref.html.in file to extract the set of a builtin function.
Right now, this file is being manually downloaded during the build phase which does not integrate well with Zig's package manager and caching system. GitHub can't provide that file as a compressed tarball and fetching the entire repository source code is too wasteful.

It would be nice if there was a way to declare it as dependency to build.zig.zon:

    .dependencies = .{
        .langref = .{
            .url = "https://raw.githubusercontent.com/ziglang/zig/5b2ee5eacc177873ce674a307a1bebdfffeeae10/doc/langref.html.in",
            // have some option here to tell Zig that the url points to a single file and does not need to be decompressed
            .hash = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",
        },
    },

[marler8997]

Here's another use case I discovered today. I'm adding a build.zig file to build VVVVVV, and it has a data.zip you can download with free assets and levels. I added the data.zip to build.zig.zon, but the game actually needs the zip file, not the extracted contents. To solve this I "re-zip" the contents and install them to the bin directory for the game to load, however, if there was a way to instead download the zip file as just a "file", this would remove some unnecessary steps.

[andrewrk]

This seems like it would be fully solved with .unpack = false, yeah?

[marler8997]

I think I'll put together a PR for this imminently as I now have a real need for it, unsupported archive formats. With the ability to disable unpacking an archive, this unlocks the ability to download an archive in any format. It can then be extracted via another dependency.

More specifically, I'm converting my win32metdata to JSON converter from C# to zig and the metadata itself is a nuget package. With this "unpack" option I can just download the nupkg file and extract it with my own extraction tool.

[castholm]

Many programs make assumptions about how to process the file based on filename extension and just inferring the filename from the URL or using it a random name or a hash might confuse such programs. It would be useful if the user could override the filename the single-file dependency is saved as. For example (continuing with the .unpack idea):

.dependencies = .{
    .template = .{
        .url = "https://example.com/headers/api-1.2.3",
        .hash = "...",
        .unpack = false,
        .filename = "api.h",
    },
},