A small terminal typing test. Odin. macOS and Linux. No runtime dependencies beyond libc.

ctype                # 25 words (default)
ctype -w 50          # 50 words
ctype -t 30          # 30-second sprint
ctype -z             # zen — free typing, raw WPM only
ctype --punct        # sentence-style punctuation, monkeytype-flavored
ctype --numbers      # mix in random numbers (10% of slots)
ctype --stats        # last 10 results
ctype --graph        # WPM trend chart

Modifiers compose with any mode: ctype -t 60 --punct --numbers.

Tab restarts the current run, Esc ends it, Ctrl-C aborts without saving.

Homebrew:

brew install en9inerd/tap/ctype

Shell script (downloads prebuilt binary to ~/.local):

curl -fsSL https://raw.githubusercontent.com/en9inerd/ctype/master/install.sh | sh

Override install prefix: CTYPE_PREFIX=/usr/local sh install.sh. Pin a version: sh install.sh v0.1.0.

Both methods install the binary and the default wordlist.

Requires Odin.

make build    # debug build
make release  # optimized build
make install  # install to ~/.local (PREFIX=... to override)

The runtime searches for a wordlist in this order:

1. --words <path> flag
2. $CTYPE_WORDS env var
3. $XDG_DATA_HOME/ctype/words.txt
4. ~/.local/share/ctype/words.txt
5. /usr/local/share/ctype/words.txt
6. /usr/share/ctype/words.txt
7. <exe>/../share/ctype/words.txt
8. ./assets/words_en.txt (repo checkout)

Format: one word per line, printable ASCII, max 30 characters. Non-conforming lines are skipped. Pipe a custom list: cat words.txt | ctype --words -.

--punct mirrors monkeytype's English punctuateWord: cascading independent rolls with lastChar guards. Sentence-start words are capitalized only (no other mark). Mid-sentence words try, in order: sentence-end (10%, sub-rolled . 80% / ? 10% / ! 10%), quote wrap (1%), single-quote wrap (1.1%), parens (1.2%), colon (1.3%), dash replace (1.4%), semicolon (1.5%), comma (20%), English contraction (50% if word matches table — are→aren't, you→you're/you'll/...). First match wins.

--numbers replaces 10% of word slots with a random 1–4 digit string, no leading zeros.

Both modifiers compose. The wordlist is never mutated; transforms happen on sampled copies.

WPM and accuracy match monkeytype:

• WPM = (correct_word_chars + correct_word_count) / 5 / minutes. Only fully-correct words contribute chars and count. In-progress word gets partial char credit if all typed chars match so far. Auto-ended last word (words mode) gets char credit but no count (no space pressed).
• Raw WPM = (all_typed_chars + spaces) / 5 / minutes.
• Accuracy = correct_keys / (correct_keys + incorrect_keys). Per-keystroke; backspacing and retyping never recovers lost accuracy.

Each completed run appends a JSON line to ~/.local/share/ctype/stats.jsonl. ctype --stats shows the last ten as a table. ctype --graph [N] draws a WPM trend chart. --reset-stats deletes the file.

odin build odin/ -out:ctype -target:linux_amd64  -o:minimal
odin build odin/ -out:ctype -target:darwin_arm64 -o:minimal

Windows is not supported (POSIX termios).

odin/
├── types.odin   shared structs and constants
├── term.odin    raw mode, signals, frame buffer, ANSI escapes
├── words.odin   wordlist parsing, sampling, punct/numbers, compaction
├── stats.odin   WPM math, JSONL i/o, --stats, --graph
├── input.odin   typing input handling, backspace, auto-end
├── render.odin  palette, viewport scrolling, draw loop
└── main.odin    argv, run loop, input dispatch