rcall

Recursive text search in file and directory names and file content — fast, minimal, script-friendly.

---

rcall config src/
rcall --case-sensitive Error logs/ src/
rcall --ignore ".venv,target" imports .

What it does

rcall walks a directory tree and reports every file or directory whose name contains the search term, and every file whose content contains it. Output is one match per line, structured for scripting.

It skips symlinked directories (no infinite loops), skips binary files (detected by byte probe), and warns about unreadable files on stderr — without stopping the search.

Install

Requires Nim 2.0 or later.

git clone https://github.com/isuzano/rcall
cd rcall
./scripts/build.sh

This produces a single binary rcall in the project root. Move it somewhere on your $PATH.

Usage

rcall TEXT [PATH ...]
rcall [--ignore list] [--no-default-ignore] [--ignore-case|--case-sensitive] TEXT [PATH ...]
rcall -h | --help

If no PATH is given, the search starts from the current directory. Multiple paths are accepted and deduplicated automatically.

Use -- to search for text that starts with a hyphen:

rcall -- --my-flag src/

Options

Flag	Description
--ignore list	Comma-separated directory names to skip
--no-default-ignore	Disable built-in ignored directories
--ignore-case	Case-insensitive matching (default)
--case-sensitive	Exact case matching
-h, --help	Show help and exit

Default ignored directories

.git, node_modules, build, dist, .cache

Disable them with --no-default-ignore. Add your own with --ignore.

Output

Every match is a tab-separated line on stdout:

path    line    kind

• line is 0 for name-based matches, or the line number for content matches
• kind is name or content

# Extract only paths
rcall config src/ | cut -f1

# Show only content matches with line numbers
rcall TODO . | awk -F'\t' '$3 == "content" { print $1 ":" $2 }'

# Count matches
rcall error logs/ | wc -l

Warnings go to stderr. The search continues past unreadable files and invalid paths.

Exit codes

Code	Meaning
0	At least one match found
1	No matches found
2	Usage or argument error

This follows the grep convention, making rcall composable in shell pipelines and if conditions.

Design

• No dependencies beyond the Nim standard library
• Reads files line by line — no full file loaded into memory
• Binary detection scans the first 4 KB (null byte → binary; >10% control characters → binary)
• termLower computed once per run, not per line
• Deduplication across overlapping roots via a typed hash set
• Deterministic output: sorted by path, then line number, then kind

See docs/DESIGN.md for module-level documentation.

License

MIT — see LICENSE.