Skip to content

README.md

Latest commit

 

History

History
82 lines (69 loc) · 4.38 KB

README.md

File metadata and controls

82 lines (69 loc) · 4.38 KB
Raw

CodeForce

Personal repo for competitive programming practice, plus some C++/Linux systems-programming experiments. (中文版: README_ch.md)

Layout

  • scratch/ — the throwaway draft for whatever problem is currently open: main.cpp / std.cpp / in.txt / out.txt. std.cpp is the local brute-force/reference solution used for stress-testing against main.cpp. All four have git update-index --skip-worktree set, so you can edit them freely and git status will never show changes. The cloud copy stays a minimal placeholder; that skip-worktree state is local only and doesn't survive a fresh clone — run ./bootstrap.sh once on a new machine to set it back up (it also wires up the pre-commit hook, see below). Feel free to drop extra one-off files here too (main2.cpp, std3.cpp, ...) when you want a quick side-by-side comparison; once you're done, just rm scratch/main2.cpp scratch/std3.cpp ... for those extras. Don't delete main.cpp/std.cpp/in.txt/out.txt themselves — they're skip-worktree, so git checkout can't bring them back; you'd have to recreate them by hand and re-run git update-index --skip-worktree.
  • template/ — algorithm templates (DSU, segment tree, Fenwick tree, Treap, Dinic, Modint, etc.), meant to be copy-pasted straight into scratch/main.cpp.
  • history/ — archived submissions, organized by platform, filenames keep their original platform prefix (e.g. cf2249A.cpp, abc466G.cpp):
    • history/codeforces/ — Codeforces rated rounds (Div/Educational Round, etc.).
    • history/atcoder/ — AtCoder.
    • history/gym/ — private Codeforces Gym/virtual-participation contests a friend set up (Bround* = Baozii Round, Zround* = Zrnstsr Round, bcup* = Baozii Cup) — not rated rounds.
    • history/dmoj/ — DMOJ.
    • history/waterloo/ — Waterloo local-contest practice: w25fall* = fall 2025, w26local* = winter 2026.
  • systems/ — C++/Linux systems-programming experiments (concurrency primitives, distributed experiments, UDP/CIDR networking experiments, etc.) — not competitive-programming code.
  • comm_test/ — local testing tool for two-program interactive ("run-twice" / communication) problems, see below.
  • .vscode/, .idea/ — personal editor state, local only, untracked from git (see .gitignore). .githooks/ is different: it's shared project infrastructure (the pre-commit hook), stays tracked and ships with every clone.

First clone

./bootstrap.sh

Points core.hooksPath at .githooks/ (the pre-commit hook refuses to commit build artifacts or files over 5MB; override with ALLOW_BINARY=1/ALLOW_BIG=1 if one genuinely belongs), and re-applies skip-worktree to scratch/main.cpp/scratch/std.cpp/scratch/in.txt/scratch/out.txt.

Day-to-day practice: run.sh

./run.sh scratch/main [in.txt] [out.txt]
./run.sh -h                     # full usage (every env-var flag, its default, example combos)
./run.sh clean                  # wipe .build/, source files untouched
  • Compiles <name>.cpp into .build/<name>, -O0 -g debug build by default. Every build artifact (binary, .dSYM, program output) lives in .build/ — nothing spills into the repo root, and none of it is tracked by git.
  • Environment-variable flags (full list in ./run.sh -h):
    • RUN=1 — run once against in.txt right after compiling, and diff against out.txt.
    • SAN=1 — turn on ASan/UBSan.
    • PERF=1 — switch to a -O3 performance build (no sanitizer).
    • DEBUG=1 — drop into lldb after compiling.
    • INTERACTIVE=1 — skip file input, run interactively instead.
    • NO_DIFF=1 — just show the program's output, skip the diff.
    • LLVM=1 — emergency fallback: compile with the standalone Homebrew LLVM instead.

Two-program interactive problems: comm_test/

Problems that run twice, where the two programs can't talk to each other directly and only communicate through interactive queries (run-twice / communication problems), don't fit run.sh's "diff two files" model — see the header comment in comm_test/run_comm.sh for why. Usage:

./comm_test/run_comm.sh scratch/sol [in.txt] [-- extra args, e.g. -v]
./comm_test/stress.sh .build/scratch/sol [iterations]   # loop random cases under stress

in.txt is the raw jury input fed to the first run (starts with first); comm_test/gen.py can generate random/edge-case inputs.