crys

🚧 Very early stage

A command-line tool that can process text files ranging from a few hundred megabytes to several gigabytes with a single command. It uses a Ruby-like syntax and is about as fast as C.

Installation

Build from source:

git clone https://github.com/kojix2/crys
cd crys
make release=1 parallel=1
make install

Q: Building from source? Why don’t you distribute pre-compiled binary files?

A: This tool only works in environments where the Crystal compiler is available. If the Crystal code cannot be compiled, Crys will not run. That’s why this approach works.

Usage

Basic form:

crys [options] 'CRYSTAL_CODE' [file ...]

Process stdin line by line:

printf 'a\nb\n' | crys -n 'puts l.upcase'

Assign the result back to l and print it:

printf 'a\nb\n' | crys -p 'l.upcase'

Auto-split input:

printf 'a:b\nc:d\n' | crys -a -d: 'puts f[1]'

Auto-split input with regex separator:

printf 'a:  b\nc:   d\n' | crys -a -d'/: +/' 'puts f[1]'

Read a full JSON document from input:

printf '{"a":1}' | crys -r json 'puts JSON.parse(ARGF)["a"].as_i'

Run setup and teardown code:

printf '1\n2\n3\n' | crys -A 'sum = 0' -n 'sum += l.to_i' -E 'puts sum'

Edit files in place:

crys -I .bak -p 'l.gsub("foo", "bar")' file.txt
crys -i -p 'l.upcase' file.txt

Inspect generated code:

crys -D -p 'l.upcase'

Filter lines with repeatable preconditions:

printf 'ok\nerror\nwarn\n' | crys -n --where 'l =~ /err|warn/' 'puts l'

Use shortcut selectors and mappers:

printf 'a\nb\n' | crys -F 'l == "a"'
printf 'a\nb\n' | crys -M 'l.upcase'

Bind split fields to names:

printf 'alice:20\nbob:30\n' | crys -a -d: -N name,age 'puts "#{name}:#{age}"'

Use header-based access:

printf 'name,age\nalice,20\n' | crys -a -d, -H -M 'row["name"]'

Aggregate quickly without boilerplate:

printf '1\n2\n3\n' | crys -S 'l.to_i'
printf 'ok\nerr\nwarn\n' | crys -W 'l =~ /err|warn/' -C
printf '1\nfoo\n3\n' | crys -W 'l =~ /^[0-9]+$/' -S 'l.to_i' -C

Options

• -n: read input line by line. Exposes l, nr, and fnr
• -p, --print: same as -n, but assigns the body result back to l and prints it
• -a, --auto-split: auto-split l into f and expose nf
• -d, --delimiter SEP: field separator for -a. Prefix with / and suffix with / to use a regex: -d'/: +/'
• -N, --names NAMES: bind split fields to variable names. Example: -N name,count
• -W, --where COND: pre-filter condition in line mode. Repeatable, combined with AND
• -M, --map EXPR: shortcut for line mode mapping (puts(EXPR))
• -F, --filter COND: shortcut for line mode filtering (puts l if COND)
• -H, --header: treat first row as header and expose row hash (requires -a)
• -S, --sum EXPR: sum expression across selected rows; exposes __crys_sum
• -C, --count: count selected rows; exposes __crys_count
• -i, --inplace: edit files in place without backup
• -I SUFFIX: edit files in place and keep backups with SUFFIX
• -r, --require LIB: add require "LIB" to the generated program. Resolution is done from CRYS_HOME
• -A, --init CODE: insert code before the main body or loop
• -E, --final CODE: insert code after the main body or loop
• -D, --dump: print the generated Crystal code and exit
• -O LEVEL: build with optimization level (0, 1, 2, 3, s, z)
• -R, --release: build with crystal build --release
• --error-trace: build with crystal build --error-trace
• --version: show tool version
• -h, --help: show help

Implicit Variables

• l: current line, always chomped
• f: split fields, only with -a
• nf: number of fields (f.size), only with -a
• nr: record number (global, counts across all files)
• fnr: per-file record number (same as nr for stdin, resets to 1 at each new file)
• path: current file path when reading files or editing in place
• row: Hash(String, String) mapped from header columns, only with --header

Dependency Resolution (CRYS_HOME)

crys builds and runs generated programs under CRYS_HOME (default: ~/.local/share/crys). When you use -r LIB, dependency resolution is performed from this directory.

Typical setup:

export CRYS_HOME="$HOME/.local/share/crys"
mkdir -p "$CRYS_HOME"
cd "$CRYS_HOME"
shards init
vi shard.yml # Add your favorite shards
shards install

Then:

printf '{"a":1}' | crys -r json 'puts JSON.parse(ARGF)["a"].as_i'

Caching

Generated programs are cached under CRYS_HOME/cache and reused when the generated code and Crystal flags are unchanged.

Constraints

• -i requires at least one file
• -M and -F cannot be combined
• -M / -F cannot be combined with explicit CRYSTAL_CODE
• -N requires -a
• -H requires -a
• -S / -C cannot be combined with explicit CRYSTAL_CODE

Development

Build:

make

Run unit tests:

make test-unit

Run integration tests:

make test-integration

Run all tests:

make test