Field guide / Rule 12 — White space
A working document collects white space the way a desk collects paper — a doubled space here, a trailing gap there, an invisible character pasted in from somewhere else. None of it shows, until it nudges a line or breaks a search. This is the cleanup we run on every file, in every language.
Proportional type only needs one space between words, so any run of them collapses to one. Spaces left dangling at the end of a line are trimmed — while a deliberate two-space line break is kept. Stacks of blank lines settle to a single blank between paragraphs, and the stray spaces after a heading’s # are tidied so the heading is read as a heading. None of it changes a word; it just removes the slack.
Text pasted from the web or another program often carries white space that isn’t a plain space at all — a fixed non-breaking space, a hair space, a zero-width character that shows as nothing yet breaks find-in-page and copy-paste. We normalise those exotic spaces back to a regular one, so what looks like a space behaves like a space. (Where a non-breaking space is genuinely wanted — binding a number to its unit — we add it deliberately; see non-breaking spaces.)
Everything here is universal — it runs whatever the document’s language. It only ever removes or normalises white space; it never touches a word, and it leaves code and its indentation alone.
Runs of spaces collapse to a single one.
No space sits before a comma or other punctuation.
A percent sign glues to its number.
Extra spaces after a heading marker are tidied.
In short
White space should be uniform and invisible — one space between words, none before punctuation, nothing exotic hiding in the gaps. We sweep all of it on the way out, in any language, without ever changing a word or touching your code.
Want to see it on your own text? Convert a document and read the cleanup report — every space we cleaned is counted there.