Whitespace and Invisible Character Cleaner

The whitespace you cannot see

A space is not always a space. Unicode defines more than a dozen characters that render as blank, and text copied out of a PDF, a Word document or a styled web page is full of them. They all survive a find-and-replace for a normal space, because they are not one.

• U+00A0, the non-breaking space. Comes from   in HTML, from Word's autoformatting, and on a Mac from Option plus the space bar. It is the single most common offender.
• U+202F, the narrow non-breaking space. French typography puts one before a colon, semicolon, question mark or exclamation mark.
• U+2000 to U+200A. En space, em space, thin space, hair space, figure space — typesetting widths that end up in text exported from layout software.
• U+3000, the ideographic space. What the space bar produces in a Japanese or Chinese input method. It is full width, so it looks like two spaces.
• U+200B, the zero-width space. Has no width at all. Some content management systems insert it inside long words so they can wrap.
• U+FEFF, the byte order mark. Belongs at the start of a file, not in the middle of a string, and shows up in both places.

The damage is always the same shape: two strings that look identical do not match. A spreadsheet lookup returns an error on a name you can see is correct. A JSON key fails to resolve. A pasted password is rejected. A CSV import puts a leading space in every field of one column.

Line endings, and why no tool in a text box can fix them

Windows ends a line with two characters, a carriage return then a line feed. Unix and macOS use the line feed alone. Classic Mac OS, before OS X, used the carriage return alone. Mixing them is what makes a file open as one enormous line in one editor and show a trailing ^M on every line in another, and what makes a diff claim every line changed when only one did.

Here is the part worth knowing: this tool cannot help you with that, and neither can any other page with a text box. HTML normalises the contents of a textarea to plain line feeds, so by the time a script reads what you pasted, the carriage returns are already gone. There is nothing left to count and nothing left to report. Line endings are a file-level problem, so fix them at file level — in your editor's line-ending setting, with dos2unix, or with git's own conversion setting.

Why collapsing spaces is sometimes the wrong move

Two of the options here are destructive in ways that are obvious in hindsight:

• Python, YAML and Makefiles. Indentation is syntax. Trimming the start of each line turns a working file into a syntax error, and a Makefile specifically requires a real tab.
• Markdown. Two spaces at the end of a line mean a hard line break. Four spaces at the start mean a code block. Both vanish when you trim and collapse.
• Fixed-width data and ASCII tables. Columns are defined by character position. Collapsing runs of spaces merges every column into one.

The rule of thumb: run this on prose. For code, turn off the line trimming and the space collapsing, keep the invisible character removal, and let your formatter handle the rest.

Removing invisible characters can break emoji

One of the characters in the invisible set is U+200D, the zero-width joiner, and it does real work. A single developer emoji is a person, a joiner and a laptop. Strip the joiner and you get two separate emoji instead of one. Family emoji fall apart the same way. If your text has emoji in it and they matter, switch that option off and clean the spaces only.

What this tool will not fix

It cannot read intent. A non-breaking space between a number and its unit, or inside a name, is deliberate typography — it is there to stop a line break splitting "10 km" in half. This tool replaces it with an ordinary space along with all the accidental ones. If you write carefully, compare before and after rather than trusting the output.

It also leaves the other invisible-difference problem alone: homoglyphs. A Cyrillic а and a Latin a are different characters that render identically, and no amount of whitespace cleaning will make them match. Same symptom, different cause, and worth ruling out when a string still refuses to compare equal after a clean.

Curly quotes, em dashes and ellipsis characters are left exactly as they are, because they are usually correct. And everything here runs synchronously in the page, so a paste of several megabytes will freeze the tab for a moment while it works. For files that size, a text editor with a regular expression search is the better tool.

Last updated September 19, 2026