tblx — binary columnar files for the terminal

tblx is the CLI + TUI of tblx (called Tablix, pronounced tab-lix): it converts CSV → TBLX through an interactive type wizard, browses .tblx files in the terminal, prints metadata, and exports back to CSV.

📖 Full documentation lives in the wiki — the complete TBLX spec, the library guide, extending TBLX, and a tour of every file.

The format itself lives in its own library, libtblx — this repo only contains the command-line tool. The website lives in tblxweb

tblx/
├── cmd/tblx/main.go             import · view · info · export
├── internal/tui/import.go       Bubble Tea wizard (↑↓ select, ←→ cycle type)
├── internal/tui/view.go         Bubble Tea viewer (frozen header, j/k, pan)
├── packaging/rpm/tblx.spec      Fedora / RHEL package
├── packaging/arch/PKGBUILD      Arch Linux package
├── wiki/                        GitHub wiki pages (see wiki/README.md)
└── samples/test.csv             demo data (includes missing values)

Build

Requires Go ≥ 1.26.

go mod tidy        # generates go.sum on first checkout
go build -o tblx ./cmd/tblx

Commands

tblx import <file.csv>            CSV → TBLX via an interactive type wizard
tblx view   <file.tblx>           scrollable table viewer
tblx info   <file.tblx>           metadata + schema
tblx export <file.tblx> [out.csv] TBLX → CSV (stdout if out.csv is omitted)

Typical Session

$ tblx import samples/test.csv    # wizard: ↑/↓ select column, ←/→ cycle
                                  #         int → float → string, enter confirms
wrote samples/test.tblx — 5 rows x 3 cols, 180 B
   1. name             string
   2. age              int
   3. score            float

$ tblx info samples/test.tblx
file     samples/test.tblx
format   TBLX (magic "TBLX")
rows     5
columns  3
size     180 B (180 bytes)
schema
   1. name             string  block 26 B
   2. age              int     block 40 B
   3. score            float   block 40 B

$ tblx view samples/test.tblx     # frozen header, j/k scroll, ←/→ pan, q quits

$ tblx export samples/test.tblx roundtrip.csv

The import wizard guesses each column's type (via libtblx.GuessTypes: int if every sample parses as int64, else float, else string) — you only touch the columns it got wrong. TBLX has no NULL: empty cells are stored as 0 / 0.0 / "".

Packaging

Fedora / RHEL (RPM)

cd packaging/rpm
rpmbuild -bb tblx.spec
sudo dnf install "$HOME/rpmbuild/RPMS/$(uname -m)/tblx-1.0.0-1.$(uname -m).rpm"

Arch Linux (pacman)

cd packaging/arch
makepkg -si

Both packages install a single statically-linked binary to /usr/bin/tblx (CGO_ENABLED=0, -trimpath -ldflags "-s -w").

Automated: every GitHub Release builds packages for you

.github/workflows/release.yml runs whenever you publish a release and attaches ready-to-install assets to it:

asset	built in	install
tblx-<ver>-1.fc41.<arch>.rpm	a fedora:41 container	sudo dnf install ./tblx-….rpm
tblx-<ver>-<arch>.pkg.tar.zst	an archlinux container (makepkg)	sudo pacman -U ./tblx-….pkg.tar.zst
tblx-<ver>-{linux,darwin}-{amd64,arm64}	static binaries	chmod +x and run
tblx-<ver>.tar.gz	source tarball	—

Verifying the format

$ tblx import samples/test.csv && xxd samples/test.tblx | head -3
00000000: 5442 4c58 0500 0000 0000 0000 0300 046e  TBLX...........n
00000010: 616d 6503 0003 6167 6501 0005 7363 6f72  ame...age...scor
00000020: 6502 0029 0000 0000 0000 0028 0000 0000  e..).......(....

The file opens with 54 42 4C 58 ("TBLX"), then row_count = 5, col_count = 3 — exactly as specified.

License

MIT