Prepare reproducible text fixtures and decorative samples safely

Problem

Text fixtures often begin as a quick request to make filler, reverse a label, scramble words, repeat a marker, show Braille, or add a dramatic effect. That request hides decisions that determine whether the result is reproducible and safe: JavaScript code units can split a displayed grapheme; word operations can destroy punctuation and line endings; unseeded random output makes screenshots and bug reports drift; repetition and combining marks can grow unexpectedly; FIGlet accepts only printable ASCII; an English Grade 1 Braille subset is not translation or accessible tactile layout; and decorative mirror, upside-down, or Zalgo text can be unreadable or irreversible. Without a written contract, a disposable visual sample can leak into production copy or a test fixture can fail for reasons unrelated to the feature under test.

When to use this

• A screenshot, visual regression test, component story, or mockup needs placeholder text that can be regenerated exactly.
• A multilingual test must reverse or shuffle grapheme, word, or line units without normalizing punctuation and whitespace.
• A terminal example or README needs a short bounded FIGlet heading while retaining a normal text heading for accessibility.
• A learning example needs to inspect a documented English Grade 1 Braille subset without claiming translation or certification.
• A disposable visual experiment needs mirror, upside-down, repetition, or Zalgo output with explicit limits and a retained plain-text source.

Steps

1. Step 1

   Classify the artifact before transforming text

   Write whether the result is a test fixture, documentation example, terminal decoration, learning sample, or production-facing copy. Name the environment that will consume it and what failure would matter. Production copy should normally remain ordinary text. Decoration is acceptable only when losing the effect, font, marks, or layout cannot hide instructions, names, safety information, controls, or the only accessible label.

2. Step 2

   Preserve and sanitize the source

   Keep an untouched source with a stable fixture name or version. Remove secrets, customer data, production identifiers, and copyrighted material that the fixture does not need. Local browser processing means the workbench does not upload the source, but clipboard history, downloaded files, screenshots, browser extensions, repositories, and test reports can still expose it. Do not normalize or remove invisible characters until you decide whether they are evidence for the test.

3. Step 3

   Choose a structural unit and locale explicitly

   Use Intl.Segmenter graphemes when combining marks and joined emoji must move together. Use word slots when punctuation, repeated spaces, and line breaks must stay where they are, and record English, Korean, or Japanese segmentation. Use logical lines when CRLF, LF, or CR separator slots matter. State whether the operation reverses all selected units, reverses each word internally, or shuffles values with separators held in place.

4. Step 4

   Freeze generation and growth controls

   For Lorem, shuffle, or Zalgo, record the unsigned 32-bit seed and every option before viewing the result. A seed gives reproducibility for the same deployed algorithm; it is not security randomness. For repetition, calculate count and separator intent and respect the 2 MiB output preflight. Keep custom separators under 256 UTF-8 bytes, Lorem quantity at 100 or less, repeat count at 10,000 or less, and Zalgo input under 32,768 code points with chaos from 1 through 10.

5. Step 5

   Apply format-specific boundaries

   Keep FIGlet input within 96 printable ASCII characters and three lines, then retain a plain heading. Treat the Braille mode as a limited English uncontracted Grade 1 subset with capital, number, Grade 1, and common punctuation signs; do not use it as translation, contracted Braille, tactile layout, math or music Braille, or certification. Treat mirror and upside-down tables as font-dependent approximations. Treat Zalgo as an accessibility stress sample and select only the mark directions the test needs.

6. Step 6

   Verify the full result and store its contract

   Run the dedicated Worker and inspect input and output UTF-8 bytes, Unicode code points, mode-specific counts, unsupported-character notices, seed, and accessibility warnings. Copy or download the full bounded output because the screen preview stops at 50,000 code points. Re-run the saved contract, compare with Diff Checker when exact fixture stability matters, and store source, mode, options, seed, browser date, expected output, and plain-text alternative together.

Example

Purpose: test whether a card preserves punctuation and two line breaks. Source: ‘one,  two!
안녕 세계
東京 テスト’. Contract: shuffle word slots, Korean segmentation, seed 42, keep the untouched source and no production data.

Save the exact shuffled output together with seed 42, mode, locale, source version, UTF-8 byte counts, browser date, and expected punctuation and newline slots. The words may change slots, while comma, double space, exclamation mark, and both line breaks remain in their original structural positions.

Common mistakes

FAQ