Field guide / Rule 09 — Accents
An é can be written two ways that look identical on the page — as one real letter, or as a plain e with a floating accent stacked on top. They read the same, but they are not the same, and the difference quietly breaks search, sorting and fonts.
There is a single, precomposed character for é — one letter, complete with its accent. But the same shape can also be built from two pieces: a plain e followed by a combining accent that the software stacks over it. On screen the two are indistinguishable. Underneath, one is a single letter and the other is a letter wearing a hat.
Text picks up the decomposed form all the time — pasted from a PDF, typed on certain keyboards, exported by some tools — and nothing looks wrong, which is exactly why it survives into the finished document.
Because the two forms are different sequences of characters, a computer treats them as different text. Search for café and a decomposed café may not match. Alphabetise a list and the odd one sorts to the wrong place. Change fonts and the stacked accent can drift off-centre, since the font was drawn expecting the single letter. None of it shows until it bites — a failed find-in-page, a name filed under the wrong letter.
We fold every decomposed accent back into its single, precomposed character, so the whole document uses the real letters throughout. It is a lossless repair — the page looks exactly the same, but search, sorting and font rendering start working again. What we will never do is guess a missing accent: if a word arrives without one, only its author knows whether it should have it, so it stays as written.
The stacked accent folds into one real letter — now searchable.
Same for the tilde, the cedilla, every diacritic Portuguese uses.
No mark is present, so there is nothing to fold — and we never add one.
In short
An accented letter should be one real character, not a letter with a mark stacked on top — the two look identical but behave differently in search, sorting and type. We quietly fold every accent already in your text into its true letter, and we never invent one that isn’t there.
Want to see it on your own text? Convert a document and read the cleanup report — every accent we folded is counted there.