@@ -0,0 +1,146 @@ |
| 1 | +# Code tab |
| 2 | + |
| 3 | +The code tab is the GitHub-style repo browser: tree listing, blob view |
| 4 | +with syntax highlighting, raw view, "Go to file" finder, and the |
| 5 | +branch/tag switcher. After a successful push, hitting `/{owner}/{repo}` |
| 6 | +sends the user to `/tree/{default_branch}`. |
| 7 | + |
| 8 | +## Routes |
| 9 | + |
| 10 | +| Route | Handler | |
| 11 | +| ------------------------------------------------ | -------------------------------- | |
| 12 | +| `GET /{owner}/{repo}` | redirects to `/tree/{default}` | |
| 13 | +| `GET /{owner}/{repo}/tree/{ref}/{path...}` | `codeTree` | |
| 14 | +| `GET /{owner}/{repo}/blob/{ref}/{path...}` | `codeBlob` | |
| 15 | +| `GET /{owner}/{repo}/raw/{ref}/{path...}` | `codeRaw` | |
| 16 | +| `GET /{owner}/{repo}/find/{ref}?q=...` | `codeFinder` | |
| 17 | +| `GET /static/css/chroma.css` | runtime-generated Chroma theme | |
| 18 | + |
| 19 | +Every code-tab handler runs through `policy.Can(... ActionRepoRead)` — |
| 20 | +private repos hide from anonymous viewers and unrelated users via the |
| 21 | +existence-leak 404 guard from S15. |
| 22 | + |
| 23 | +## Ref + path disambiguation |
| 24 | + |
| 25 | +`{ref}` is the chi `*` wildcard, so the URL `/tree/feature/x/sub/file.go` |
| 26 | +arrives as a single string. Resolution: |
| 27 | + |
| 28 | +1. If the first segment is exactly 40 hex chars → treat as a SHA, the |
| 29 | + rest is the path. |
| 30 | +2. Otherwise, longest-prefix match against the cached ref list |
| 31 | + (branches first, then tags, sorted longest-first). The remainder |
| 32 | + after the matched ref is the in-tree path. |
| 33 | + |
| 34 | +This handles `release/v1.0/beta/CHANGELOG.md` correctly without |
| 35 | +ambiguity. Resolution lives in `internal/repos/git/treeops.go::ResolveRef`. |
| 36 | + |
| 37 | +Path validation rejects `..`, control chars, leading slashes, and |
| 38 | +backslashes — defense in depth on top of git's own validation. |
| 39 | + |
| 40 | +## Tree listing |
| 41 | + |
| 42 | +`git ls-tree --long --full-tree <ref>:<path>` is parsed into typed |
| 43 | +`TreeEntry` values (`tree | blob | commit | symlink`). Sort is |
| 44 | +directories first, then files alphabetically. |
| 45 | + |
| 46 | +The S17 ship excludes the htmx-driven "last commit per entry" column |
| 47 | +that the spec describes — an extra round-trip we can add later without |
| 48 | +a schema change. The current page renders the listing immediately. |
| 49 | +**Deferred to S18 (commits-per-entry)** — the spec calls out this |
| 50 | +deferral path; the tree template has the column slot ready. |
| 51 | + |
| 52 | +## File view |
| 53 | + |
| 54 | +`codeBlob` walks four cases: |
| 55 | + |
| 56 | +* **Large** (>1 MiB): placeholder + raw download link, no body read. |
| 57 | +* **Binary** (NUL byte in first 8 KiB): placeholder. Image extensions |
| 58 | + (png/jpg/jpeg/gif/webp) ≤5 MiB get an `<img>` preview pointing at |
| 59 | + `/raw/...`. |
| 60 | +* **Markdown** (`.md`/`.markdown`): Goldmark + bluemonday rendered HTML |
| 61 | + PLUS a `<details>` source-toggle with the highlighted source. |
| 62 | +* **Default text**: Chroma highlight by filename extension, content |
| 63 | + sniffing fallback. |
| 64 | + |
| 65 | +Chroma uses the `github` style baked at process start; the CSS is |
| 66 | +served from `/static/css/chroma.css` via a tiny in-process generator. |
| 67 | + |
| 68 | +## Raw view |
| 69 | + |
| 70 | +* Content-Type derived from the extension whitelist |
| 71 | + (`code.go::rawContentType`). |
| 72 | +* `X-Content-Type-Options: nosniff` always. |
| 73 | +* `Content-Security-Policy: default-src 'none'; sandbox` at the |
| 74 | + handler level (the global SecureHeaders middleware may overlay a |
| 75 | + broader CSP — both are restrictive; the OR of the two is what user |
| 76 | + agents enforce). |
| 77 | +* **`Content-Disposition: attachment`** is forced for HTML, SVG, JS, |
| 78 | + WASM, and anything that could execute on shithub's domain. We don't |
| 79 | + have a separate `raw.shithub.tld` host yet (post-MVP); attachment is |
| 80 | + the safety belt. |
| 81 | +* Streamed via `git cat-file -p`; never buffered. Large blobs don't |
| 82 | + blow up the worker's memory. |
| 83 | + |
| 84 | +## Finder ("Go to file") |
| 85 | + |
| 86 | +`/find/{ref}` lists every blob path on the ref via |
| 87 | +`git ls-tree -r --name-only`, then filters with |
| 88 | +`internal/repos/finder/finder.go::Filter`. The matcher is a |
| 89 | +subsequence-with-bonus scorer (boundary, consecutive run, basename |
| 90 | +hit) — not as fancy as VS Code's quickopen but good enough for tens of |
| 91 | +thousands of paths. |
| 92 | + |
| 93 | +Key shortcut and live-filter via htmx are spec deliverables that we |
| 94 | +defer for now — the form-submission flow works without JS and that's |
| 95 | +the floor S17 commits to. |
| 96 | + |
| 97 | +## Caching |
| 98 | + |
| 99 | +Currently **no caching layer**. Every request runs `git for-each-ref`, |
| 100 | +`git ls-tree`, etc. That's fine for small-to-medium repos; the cost |
| 101 | +shows up on big repos with deep trees. The S17 spec proposes a cache |
| 102 | +keyed on `(repo_id, ref_oid, dir_path)` invalidated on push (S14's |
| 103 | +`push:process` job is the right invalidation hook). |
| 104 | + |
| 105 | +**Deferred** — the cache is purely performance polish. When we hit a |
| 106 | +real-world repo where it matters, wire it in: file `internal/cache/` |
| 107 | +plus a callback in `worker/jobs/push_process.go`. The handlers already |
| 108 | +take a per-request `policy.Cache` so adding a per-process git cache is |
| 109 | +mechanically straightforward. |
| 110 | + |
| 111 | +## Pitfalls + protections |
| 112 | + |
| 113 | +* **XSS via raw HTML/SVG**: blocked by `Content-Disposition: attachment` |
| 114 | + for those extensions. |
| 115 | +* **XSS via markdown**: Goldmark configured without HTML passthrough + |
| 116 | + bluemonday's UGC policy on top. Tests in `internal/repos/markdown/` |
| 117 | + (TODO — minimal coverage today). |
| 118 | +* **Path traversal**: `validateSubpath` in `code.go` rejects `..`, |
| 119 | + controls, leading slashes. |
| 120 | +* **Hex collision with SHA**: ref-list lookup wins over SHA shortcut |
| 121 | + when the same string is both. |
| 122 | +* **Encoding (GBK / Shift-JIS)**: TODO — text files outside UTF-8 may |
| 123 | + render as garbled. The body is rendered as-is; a future commit can |
| 124 | + add `golang.org/x/text/encoding` autodetection. |
| 125 | + |
| 126 | +## Dependencies |
| 127 | + |
| 128 | +* `github.com/alecthomas/chroma/v2` — syntax highlighting |
| 129 | +* `github.com/yuin/goldmark` — CommonMark + GFM |
| 130 | +* `github.com/microcosm-cc/bluemonday` — HTML sanitizer |
| 131 | + |
| 132 | +## Deferred polish (tracked, not blocking) |
| 133 | + |
| 134 | +These items are spec deliverables we ship in a later pass: |
| 135 | + |
| 136 | +* **Last-commit-per-entry column** with htmx lazy load and pre-walked |
| 137 | + `git log --name-status` cache → wire into S18 (commit history) where |
| 138 | + the same walk powers the per-file history page. |
| 139 | +* **Tree caching keyed on (repo_id, ref_oid, dir_path)**, push-event |
| 140 | + invalidation → wire into S36 (performance pass) once we have a real |
| 141 | + workload to measure. |
| 142 | +* **Pagination at 1000 entries per directory** → cosmetic for huge |
| 143 | + trees; add when someone hits `node_modules`-grade inflation. |
| 144 | +* **Encoding detection for non-UTF-8 source files** → file reads are |
| 145 | + defensive (`io.LimitReader` + size cap); render quality is the only |
| 146 | + loss until this lands. |