Fun facts

"Ti" stands for Titanium

From the README:

TiDB (/'taɪdiːbi:/, "Ti" stands for Titanium) is an open-source, cloud-native, distributed SQL database…

The naming convention extends across the family: TiKV (Titanium Key-Value), TiFlash (Titanium Flash, the columnar replica), TiDB Lightning, TiCDC, etc.

The 8 MB single source file

pkg/parser/parser.go is 7,952,510 bytes of generated yacc tables — almost 8 MB in one file. It is regenerated from pkg/parser/parser.y (~421 KB of grammar). Most editors will think hard before opening it; many tooling integrations (gopls, gocopy) explicitly skip it.

A few honourable mentions in the "very large generated file" category:

Bytes	File
7.95 MB	pkg/parser/parser.go
2.16 MB	pkg/util/collate/ucadata/unicode_0900_ai_ci_data_generated.go
752 KB	pkg/util/collate/ucadata/unicode_ci_data_generated.go
631 KB	pkg/expression/builtin_time_vec_generated_test.go
437 KB	pkg/util/collate/gbk_chinese_ci_data.go

The bot top-committer

Across all 28k commits, Ti Chi Robot is the most prolific committer with 4,089 commits — almost twice as many as the second-place bot (ti-srebot, 2,020) and far more than any human. Both bots are part of the project's infrastructure; they cherry-pick approved PRs into release branches, autoland after CI passes, and clean up generated files. The most prolific human committer is Weizhen Wang at 1,819 commits.

The first commit

Commit 0d6f270068 on 2015-09-06 with the magnificently understated message:

Add files

13 minutes later (ec1fe126ee):

Update README.md

The Merge of PR #1 (a README patch) followed within the same hour. The codebase has grown by 28,000 commits and 10½ years since then.

TODO archaeology

pkg/ contains 2,361 instances of TODO, FIXME, or HACK. That's roughly one marker for every 13 source files; not bad for a codebase of this size, but enough that there's a cleanup target there for anyone who wants one.

Skills the repo expects you to know

The .agents/skills/ directory contains seven repo-local "skills" — operational playbooks that contributors and AI agents are expected to follow:

• tidb-bazel-prepare-gate
• tidb-failpoint-test-runner
• tidb-integrationtest-recorder
• tidb-realtikv-runner
• tidb-test-diff-triage
• tidb-test-guidelines
• tidb-verify-profile

The repo also defines three verification profiles (WIP, Ready, Heavy). You must run Ready before claiming a task is "done" or "ready for review", and the trigger phrases for Ready are listed in AGENTS.md. The bar for declaring victory is set carefully and explicitly.

Three commits per hour

The 489 commits in the last 90 days work out to about 5.4 commits per day, or roughly 3 commits per hour during US business hours if you assume even spread (which it isn't — the project has a strong APAC contributor base and sees activity around the clock).