Text Formatter

Why a chain instead of a button

Real text tasks are almost never a single transform. Cleaning a list pasted from a spreadsheet usually means trimming every line, dropping the blank ones, removing repeats and only then sorting. With independent buttons, each click throws away the previous result, and the output has to be pasted back into the input by hand at every stage.

Here the queue is the document. It holds the order, it holds each step's parameters, and it can be saved under a name to repeat tomorrow. Because every step has its own switch, you can isolate the effect of one of them without dismantling the queue: switch it off, look at the result, switch it back on.

• Order matters: removing accents after generating the slug does nothing, before does everything.
• A disabled step stays in the queue with its parameters intact.
• Undo and redo act on the chain, not on the text you typed.
• Presets store the whole chain, parameters included.

What each group of operations does

Case covers uppercase, lowercase, title case (first letter of each word), sentence case (first letter after a period, exclamation mark or question mark), alternating case and case inversion. Spaces covers cleanup: collapsing repeated spaces and tabs, trimming the ends of each line, removing blank lines and converting between tabs and spaces, where the spaces-to-tabs conversion touches only the indentation so it cannot damage running text.

Lines treats the list as a structure: sorting by letter in both directions, sorting by the first number on each line (which differs from sorting as text, where "10" would come before "2"), reversing the order, removing duplicates with or without case sensitivity, numbering with alignment, and shuffling from a seed so the result is reproducible.

Transform handles encodings: base64 (UTF-8 aware and also accepting the base64url variant), URL percent-encoding, HTML escaping and unescaping, JSON string literals, pretty or minified JSON, slug, accent removal and character reversal. Wrapping reflows the text: wrap at N columns without breaking long words, join every line with a separator, and undo the line breaks inside a paragraph while keeping the blank line that separates one paragraph from the next.

Extract, replace, and the guard against freezing the page

The two regex steps solve what no fixed operation can: extracting every match of a pattern (one per line, or with the separator you choose) and replacing using capture groups, where $1 and $2 point at the parentheses in the pattern, $& at the whole match and $<name> at a named group. Syntax errors are reported with the regex engine's own message and the step number.

One guard is worth explaining. Certain patterns with a nested quantifier, such as (a+)+, can fall into exponential backtracking and freeze the tab. In the browser, a regex execution already under way cannot be interrupted synchronously, so capping the time afterwards would not help. That is why the defense happens beforehand: patterns with a nested quantifier are refused with a warning, the text is capped at two hundred thousand characters for regex steps, and a time and match budget cuts execution between one match and the next.

Statistics and difference

Character, word, line and byte counts are shown for both input and output, along with the change between them. Characters are counted by grapheme, meaning what the eye calls one character: a national flag emoji and an emoji with a skin-tone modifier count as one, not as two or three, which is what a naive count returns. Bytes use UTF-8 encoding, which is the number that matters when there is a size limit on the other side.

The difference tab compares input against output line by line, with word-level highlighting inside changed lines, using the same engine as the Text Comparison tool. It answers the question every long chain raises: what exactly did that step touch?