20 file formats, every one generated at an exact size

Each of these is a real file of that format. It opens in the software that owns it, and it is exactly the number of bytes you asked for. None of them is padded zeros with an extension bolted on.

Format	Extension	Smallest file	Fidelity	Checked against
`bmp`	`.bmp`	58	full	pillow
`csv`	`.csv`	117	full	python-csv
`docx`	`.docx`	1227	full	libreoffice
`gif`	`.gif`	41	full	pillow
`html`	`.html`	118	full	python-html
`ico`	`.ico`	70	full	pillow
`jpg`	`.jpg`	602	full	pillow
`json`	`.json`	219	full	node-json
`log`	`.log`	155	full	not applicable
`md`	`.md`	0	full	not applicable
`pdf`	`.pdf`	3415	full	pdftotext
`png`	`.png`	73	full	pillow
`pptx`	`.pptx`	4826	full	libreoffice
`svg`	`.svg`	194	full	inkscape
`targz`	`.tar.gz`	9793	full	7z
`txt`	`.txt`	0	full	not applicable
`wav`	`.wav`	98	full	ffprobe
`xlsx`	`.xlsx`	1735	full	libreoffice
`xml`	`.xml`	264	full	python-xml
`zip`	`.zip`	8382	full	7z

What the columns mean

Smallest file

The fewest bytes this tool will accept for that format, including the label it writes inside the file. Ask for less and you get an error naming the floor and the reason for it, never a file of the wrong size.
Fidelity

How complete the file is. full means a reader that really parses the format accepts it, not merely that the extension matches.
Checked against

The independent reader that opens every generated file before the format ships - a separate implementation, not our own code marking its own homework.

Every format also repeats to the byte: the same recipe and seed produce identical files on any machine, which is what makes a recipe safe to commit in place of the fixtures themselves.

Settings each format accepts

Most formats take settings of their own - image dimensions, JPEG quality, PDF page count, rows and columns in a spreadsheet, how many entries go inside an archive. Set them with --set key=value on the command line, or under properties: in a recipe.

tfg generate --format jpg --size 500kb --set width=1920 --set height=1080 --set quality=85

Format	Setting	Accepts
`bmp`	`width`	1 - 20000 pixels
	`height`	1 - 20000 pixels
`docx`	`paragraphs`	1 - 50000 paragraphs
`gif`	`width`	1 - 20000 pixels
	`height`	1 - 20000 pixels
`ico`	`width`	1 - 256 pixels
	`height`	1 - 256 pixels
	`embed`	bmp, png
`jpg`	`width`	1 - 20000 pixels
	`height`	1 - 20000 pixels
	`quality`	1 - 100
`pdf`	`pages`	1 - 5000
	`page_size`	a3, a4, a5, legal, letter
`png`	`width`	1 - 20000 pixels
	`height`	1 - 20000 pixels
`pptx`	`slides`	1 - 500 slides
`targz`	`entries`	0 - 10000
	`entry_format`	text
	`entry_size`	a size such as 2mb
`wav`	`sample_rate`	8000 - 192000 hertz
	`bit_depth`	8, 16, 24, 32
	`channels`	1 - 8
	`content`	noise, silence, sweep, tone
`xlsx`	`rows`	1 - 200000 rows
	`columns`	1 - 64 columns
`zip`	`entries`	0 - 10000
	`entry_format`	text
	`entry_size`	a size such as 2mb

A value outside what a setting accepts is refused with a message naming the setting, the allowed range and what to use instead. An unknown setting is an error too, never a silent default - a typo accepted in silence gives a file with the wrong settings and an hour spent wondering why the test passes when it should not.

Run tfg formats <id> to see exactly what one format accepts on the build you have.

Archives hold real files

targz and zip can be filled with entries rather than left as an empty shell. A generated archive genuinely contains the documents it claims to contain, so anything that unpacks it during a test finds real files inside.

targets:
  - id: bundle
    format: zip
    contains:
      - format: txt
        count: 200
        size: 4kb

Formats that are not here yet

7z, tiff, webp, mp3 and mp4 are planned and not built. They are absent from the list above rather than present and half working, and asking for one is an error saying the format is unknown.

If you need one of them, or a format that is not on either list, the issue tracker is where that gets decided.