Rewrite in Zig by jstrieb · Pull Request #144 · jstrieb/github-stats

jstrieb · 2026-03-29T05:58:25Z

Before merging, I will need to do some history revision to clean things up. One of the changes on this branch moves the generated files to their own branch, and I will need to manually fix some stuff up in accordance with that.

lsnow99 · 2026-03-29T22:16:55Z

src/argparse.zig

@@ -0,0 +1,230 @@
+const std = @import("std");
+
+// Since parse is the only public function, these variables can be set there and


lsnow99 · 2026-03-29T22:19:01Z

src/argparse.zig

+    }
+}
+
+fn strip_optional(T: type) type {


Why use snake case for these?

lsnow99 · 2026-03-29T22:25:47Z

src/argparse.zig

+                        const value = try a.dupe(u8, entry.value_ptr.*);
+                        defer a.free(value);
+                        @field(result, field.name) = value.len > 0 and
+                            !std.ascii.eqlIgnoreCase(value, "false");


didn't know about this function, nice

lsnow99 · 2026-03-29T22:27:21Z

src/argparse.zig

+) !void {
+    inline for (@typeInfo(T).@"struct".fields, seen) |field, *seen_field| {
+        if (!seen_field.*) {
+            if (field.defaultValue()) |default| {


ah nice, I only ever used default_value_ptr, this is better

lsnow99 · 2026-03-29T22:28:36Z

src/argparse.zig

+                    .pointer => @field(
+                        result,
+                        field.name,
+                    ) = if (default) |p| try allocator.dupe(u8, p) else null,


Why bother dupe-ing?

lsnow99 · 2026-03-29T22:32:03Z

src/argparse.zig

+                const flag_version = try allocator.dupe(u8, field.name);
+                defer allocator.free(flag_version);
+                std.mem.replaceScalar(u8, flag_version, '_', '-');


A comptime version of dupe would be very handy. I basically made one in my argparser

lsnow99 · 2026-03-29T22:32:48Z

src/argparse.zig

+                try stdout.print("--{s}\n", .{flag_version});
+            },
+            else => {
+                const flag_version = try allocator.dupe(u8, field.name);


did you mean to call it flag_ in this else branch?

lsnow99 · 2026-03-29T22:35:15Z

src/argparse.zig

+    var env = try std.process.getEnvMap(a);
+    defer env.deinit();
+    var iterator = env.iterator();
+    while (iterator.next()) |entry| {


interesting, I did the reverse where I just looped through each field, generated the env key and did env.get() on it

might be slightly simpler

lsnow99 · 2026-03-29T22:39:43Z

src/http_client.zig

+    const arena = try allocator.create(std.heap.ArenaAllocator);
+    arena.* = std.heap.ArenaAllocator.init(allocator);


Why not var arena = std.heap.ArenaAllocator.init(allocator);?

lsnow99 · 2026-03-29T23:02:40Z

src/http_client.zig

+    }
+
+    var writer = try std.Io.Writer.Allocating.initCapacity(
+        self.arena.allocator(),


Is an arena really the right model here? I imagine you'll make a lot of http requests whose responses are fairly disposable, but whose memory footprint will grow the arena over time without freeing anything until very late in the program lifespan.

lsnow99 · 2026-03-29T23:03:24Z

src/http_client.zig

+        .extra_headers = extra_headers,
+    }) catch |err| switch (err) {
+        error.HttpConnectionClosing => {
+            // Handle a Zig HTTP bug where keep-alive connections are closed by


unfortunate, but the workaround makes sense

I imagine this (and perhaps other parts of the code) prevents you from running multiple requests in parallel

lsnow99 · 2026-03-29T23:06:15Z

src/http_client.zig

+    );
+    errdefer writer.deinit();
+    const status = (try (self.client.fetch(.{
+        .location = .{ .url = url },


shouldn't you specify .method = POST here?

lsnow99 · 2026-03-29T23:06:42Z

src/http_client.zig

+        .payload = body,
+        .headers = headers,
+    }) catch |err| switch (err) {
+        error.HttpConnectionClosing => {


personally I would be making a private inner version of get/post since these two functions are so similar

lsnow99 · 2026-03-29T23:09:55Z

src/main.zig

+        if (self.api_key) |s| allocator.free(s);
+        if (self.json_input_file) |s| allocator.free(s);
+        if (self.json_output_file) |s| allocator.free(s);
+        if (self.excluded_repos) |s| allocator.free(s);
+        if (self.excluded_langs) |s| allocator.free(s);
+        if (self.overview_output_file) |s| allocator.free(s);
+        if (self.languages_output_file) |s| allocator.free(s);
+        if (self.overview_template) |s| allocator.free(s);
+        if (self.languages_template) |s| allocator.free(s);


could always do an inline for on the fields :)

lsnow99 · 2026-03-29T23:40:03Z

src/main.zig

+
+fn overview(
+    arena: *std.heap.ArenaAllocator,
+    stats: anytype,


it's interesting, you seem to prefer anonymous structs much more than me. I would probably have a single Overview struct instead of accepting anytype. I generally think type-first which could be seen as a limitation in languages where I can't annotate as heavily as I may want.

lsnow99 · 2026-03-29T23:45:11Z

src/main.zig

+            \\
+        , .{ (i + 1) * 150, color orelse "#000", language, percent });
+    }
+    // Vulnerable to template injection. In practice, this should never happen.


Quickly looked through the possibilities, seems as though nobody but the author can really affect the contents here in a meaningful way.

lsnow99 · 2026-03-29T23:46:59Z

src/main.zig

+            try list.append(allocator, pattern);
+        }
+        break :excluded try list.toOwnedSlice(allocator);
+    } else null;


Why do you need the else null?

lsnow99 · 2026-03-29T23:48:19Z

src/main.zig

+    } else null;
+    defer if (excluded_langs) |excluded| allocator.free(excluded);
+
+    var stats: Statistics = undefined;


why use the = undefined pattern here but the labeled breaks above?

lsnow99 · 2026-03-29T23:51:28Z

src/main.zig

+) ![]const u8 {
+    const a = arena.allocator();
+    var out_data = template;
+    // Vulnerable to template injection. In practice, this should never happen.


Another idea is to just use std.fmt.print. So convert your templates to use {[lang_list]s} syntax and you can just do std.fmt.print and pass the stats struct directly

lsnow99 · 2026-03-29T23:55:56Z

src/main.zig

+        if (std.mem.eql(u8, path, "-"))
+            std.fs.File.stdin()
+        else
+            try std.fs.cwd().openFile(path, .{});


I like this formatting, I wonder if I can get zls to do it

lsnow99 · 2026-03-29T23:57:30Z

src/statistics.zig

@@ -0,0 +1,558 @@
+const std = @import("std");


GitHub, I'm coming for you.

lsnow99 · 2026-03-30T00:05:24Z

src/statistics.zig

+fn get_basic_info(
+    client: *HttpClient,
+    allocator: std.mem.Allocator,
+) !struct { []u32, []const u8, ?[]const u8 } {


personally it would bother me not having these named for the caller

lsnow99 · 2026-03-30T01:17:41Z

src/statistics.zig

+                item.delay +=
+                    std.crypto.random.intRangeAtMost(i64, 2, item.delay);
+                item.delay = @min(item.delay, 600);


I like this :)

lsnow99 · 2026-03-30T01:28:52Z

src/statistics.zig

+            .ok => {},
+            .accepted => {


this works out pretty nicely

lsnow99 · 2026-03-30T01:42:17Z

src/statistics.zig

+
+    const limit = 100;
+    if (stats.commitContributionsByRepository.len >= limit) {
+        for (&[_]usize{ 2, 3 }) |factor| {


hmm so this requires that the initial caller passes a number of months whose prime factors are 2 and 3, right?

lsnow99 · 2026-03-30T01:44:24Z

src/statistics.zig

+                }
+                return;
+            }
+        } else {


so this would be hit when months = 1 and the limit is still exceeded? Interesting, I guess this works

lsnow99 · 2026-03-30T01:49:13Z

src/statistics.zig

+                    errdefer {
+                        for (0..i, repository.languages.?) |_, l| {
+                            context.allocator.free(l.name);
+                            if (l.color) |c| context.allocator.free(c);
+                        }
+                    }


[question] what if you hit an error on line 374? This errdefer will have gone out of scope, right? so these frees don't run in that case?

lsnow99 · 2026-03-30T01:50:02Z

src/statistics.zig

+                .{ .ignore_unknown_fields = true },
+            )).count;
+        } else {
+            std.log.info(


Would expect this to be a warn

lsnow99 · 2026-03-30T01:51:08Z

src/statistics.zig

+        }
+        allocator.free(result.repositories);
+    }
+    std.sort.pdq(Repository, result.repositories, {}, struct {


haven't had use for any of the stdlib sorting functions yet, but this is good to know about

lsnow99 · 2026-03-30T02:05:10Z

src/main.zig

+        }
+    }{ .values = aggregate_stats.languages.values() });
+
+    {


why is this in a bock?

jstrieb added 30 commits February 11, 2026 20:36

Add empty Zig project

da5679d

Set up logging

6a2e061

Make an HTTP request

64503cc

Wrap HTTP client with nicer methods

6ee8c1b

Arena allocate request bodies

f893cb4

Simplify get and post args

50bb92e

Move HTTP client into separate file

73344b9

Pull GitHub contribution years from GraphQL API

8623285

Pull initial repo stats

322b0bd

Minor refactor

b758e3a

Pull repo views

5974224

Add basic logging

a6192d5

Get lines changed

2b9e70b

Deallocate correctly on errors in get_repos

b6ef8ff

Work around keep alive timeouts

312e45d

Tweak API request delay values

82c4407

Partial refactor to split up big function

ed4cca8

Refactor statistics struct into separate file

e07c288

Use probabilistic exponential backoff with jitter

ce45bf2

Fix keep alive failure retry logic once and for all

4c2a3a3

Refactor lines changed into separate function

adf437d

Get contribution lines early for faster runtime

f5bbf6a

Get username only once

871a2f8

Add very basic arg and env parsing

729d421

Fix memory leaks

1520a6c

Print usage

1713e8d

Parse different types of struct fields

68cd472

Fix free_field bug, split long lines

6aca7f1

Print raw data to stdout or a file based on CLI

7b1d9d2

Refactor parts of arg parsing into helpers

0675cea

lsnow99 reviewed Mar 29, 2026

View reviewed changes

src/argparse.zig

}

}

fn strip_optional(T: type) type {

Copy link
Copy Markdown

Collaborator

lsnow99 Mar 29, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why use snake case for these?

lsnow99 reviewed Mar 29, 2026

View reviewed changes

src/statistics.zig

@@ -0,0 +1,558 @@

const std = @import("std");

Copy link
Copy Markdown

Collaborator

lsnow99 Mar 29, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

GitHub, I'm coming for you.

lsnow99 reviewed Mar 30, 2026

View reviewed changes

src/statistics.zig

Comment on lines +512 to +513

.ok => {},

.accepted => {

Copy link
Copy Markdown

Collaborator

lsnow99 Mar 30, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this works out pretty nicely

lsnow99 reviewed Mar 30, 2026

View reviewed changes

src/main.zig

}

}{ .values = aggregate_stats.languages.values() });

{

Copy link
Copy Markdown

Collaborator

lsnow99 Mar 30, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why is this in a bock?

		@@ -0,0 +1,230 @@
		const std = @import("std");

		// Since parse is the only public function, these variables can be set there and

		const arena = try allocator.create(std.heap.ArenaAllocator);
		arena.* = std.heap.ArenaAllocator.init(allocator);

Conversation

jstrieb commented Mar 29, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants