//! Build-time generator for the web LOOK filesystem.
//!
//! Prefer Git's tracked-plus-untracked/nonignored view so a new source works in
//! the browser before it is staged, while ignored build/cache/vendor files can
//! never leak into the bundle. A source archive should still be buildable
//! without Git, so a conservative directory walk is the fallback.
const std = @import("std");

pub fn main(init: std.process.Init) !void {
    const gpa = init.gpa;
    const io = init.io;
    const args = try init.minimal.args.toSlice(init.arena.allocator());
    if (args.len != 3) return error.BadArgs;

    const root = args[1];
    const output_path = args[2];
    const paths = (try gitWorkingTreeZigFiles(gpa, io, root)) orelse try walkedZigFiles(gpa, io, root);
    defer {
        for (paths) |path| gpa.free(path);
        gpa.free(paths);
    }
    std.mem.sort([]const u8, paths, {}, lessThanPath);

    var generated: std.Io.Writer.Allocating = .init(gpa);
    defer generated.deinit();
    try generated.writer.writeAll(
        \\// Generated by tools/embed_zig_sources.zig. Do not edit.
        \\pub const Source = struct { path: []const u8, contents: []const u8 };
        \\pub const all = [_]Source{
        \\
    );

    for (paths) |path| {
        const full_path = try std.Io.Dir.path.join(gpa, &.{ root, path });
        defer gpa.free(full_path);
        // Cached entries can still name a file moved or deleted without
        // staging it—an ordinary mid-edit state, and in a jj-colocated repo the
        // normal one until the change is exported. A path no longer on disk is
        // simply not a source to embed; failing the whole web build over it
        // means an unrelated refactor breaks a shell it never touched.
        const contents = std.Io.Dir.cwd().readFileAlloc(io, full_path, gpa, .limited(64 * 1024 * 1024)) catch |err| switch (err) {
            error.FileNotFound => continue,
            else => return err,
        };
        defer gpa.free(contents);
        try generated.writer.print("    .{{ .path = \"{f}\", .contents = \"{f}\" }},\n", .{
            std.zig.fmtString(path),
            std.zig.fmtString(contents),
        });
    }
    try generated.writer.writeAll("};\n");

    const output = if (std.Io.Dir.path.isAbsolute(output_path))
        try std.Io.Dir.createFileAbsolute(io, output_path, .{})
    else
        try std.Io.Dir.cwd().createFile(io, output_path, .{});
    defer output.close(io);
    try output.writeStreamingAll(io, generated.written());
}

/// Returns null only when Git cannot supply the working-tree source list.
/// `--cached --others --exclude-standard` includes a newly-created module
/// before it is staged but retains Git's ignore rules. Contents are read later;
/// Git supplies only the path filter, never historical blobs.
fn gitWorkingTreeZigFiles(
    gpa: std.mem.Allocator,
    io: std.Io,
    root: []const u8,
) !?[][]const u8 {
    const result = std.process.run(gpa, io, .{
        .argv = &.{
            "git",      "-C",       root,                 "ls-files", "-z",
            "--cached", "--others", "--exclude-standard", "--",       "*.zig",
        },
        .stdout_limit = .limited(16 * 1024 * 1024),
        .stderr_limit = .limited(1024 * 1024),
    }) catch return null;
    defer gpa.free(result.stdout);
    defer gpa.free(result.stderr);
    switch (result.term) {
        .exited => |code| if (code != 0) return null,
        else => return null,
    }

    var count: usize = 0;
    var count_it = std.mem.splitScalar(u8, result.stdout, 0);
    while (count_it.next()) |raw_path| {
        if (raw_path.len == 0 or !std.mem.endsWith(u8, raw_path, ".zig")) continue;
        count += 1;
    }

    const paths = try gpa.alloc([]const u8, count);
    errdefer gpa.free(paths);
    var index: usize = 0;
    errdefer for (paths[0..index]) |path| gpa.free(path);
    var it = std.mem.splitScalar(u8, result.stdout, 0);
    while (it.next()) |raw_path| {
        if (raw_path.len == 0 or !std.mem.endsWith(u8, raw_path, ".zig")) continue;
        paths[index] = try normalizedPath(gpa, raw_path);
        index += 1;
    }
    std.debug.assert(index == paths.len);
    return paths;
}

fn walkedZigFiles(
    gpa: std.mem.Allocator,
    io: std.Io,
    root: []const u8,
) ![][]const u8 {
    var count: usize = 0;
    {
        var dir = if (std.Io.Dir.path.isAbsolute(root))
            try std.Io.Dir.openDirAbsolute(io, root, .{ .iterate = true })
        else
            try std.Io.Dir.cwd().openDir(io, root, .{ .iterate = true });
        defer dir.close(io);
        var walker = try dir.walk(gpa);
        defer walker.deinit();

        while (try walker.next(io)) |entry| {
            if (entry.kind == .directory and ignoredFallbackDir(entry.path)) {
                walker.leave(io);
                continue;
            }
            if ((entry.kind != .file and entry.kind != .sym_link) or !std.mem.endsWith(u8, entry.path, ".zig")) continue;
            count += 1;
        }
    }

    const paths = try gpa.alloc([]const u8, count);
    errdefer gpa.free(paths);
    var index: usize = 0;
    errdefer for (paths[0..index]) |path| gpa.free(path);
    var dir = if (std.Io.Dir.path.isAbsolute(root))
        try std.Io.Dir.openDirAbsolute(io, root, .{ .iterate = true })
    else
        try std.Io.Dir.cwd().openDir(io, root, .{ .iterate = true });
    defer dir.close(io);
    var walker = try dir.walk(gpa);
    defer walker.deinit();

    while (try walker.next(io)) |entry| {
        if (entry.kind == .directory and ignoredFallbackDir(entry.path)) {
            walker.leave(io);
            continue;
        }
        if ((entry.kind != .file and entry.kind != .sym_link) or !std.mem.endsWith(u8, entry.path, ".zig")) continue;
        paths[index] = try normalizedPath(gpa, entry.path);
        index += 1;
    }
    std.debug.assert(index == paths.len);
    return paths;
}

fn ignoredFallbackDir(path: []const u8) bool {
    var parts = std.mem.tokenizeAny(u8, path, "/\\");
    while (parts.next()) |part| {
        if (std.mem.eql(u8, part, ".git") or
            std.mem.eql(u8, part, ".zig-cache") or
            std.mem.eql(u8, part, "zig-cache") or
            std.mem.eql(u8, part, "zig-out") or
            std.mem.eql(u8, part, "zig-pkg")) return true;
    }
    return false;
}

fn normalizedPath(gpa: std.mem.Allocator, path: []const u8) ![]u8 {
    const copy = try gpa.dupe(u8, path);
    if (std.Io.Dir.path.sep != '/')
        std.mem.replaceScalar(u8, copy, std.Io.Dir.path.sep, '/');
    return copy;
}

fn lessThanPath(_: void, a: []const u8, b: []const u8) bool {
    return std.mem.order(u8, a, b) == .lt;
}