Typographer: guillemets, dashes, nonbreaking spaces

How to use it

1. Choose the input type. Plain text is edited in full. In HTML mode, the typographer parses the markup and touches only text nodes: tags, attribute values, existing entities, and the contents of script, style, pre, code, and textarea remain unchanged. In Markdown, fenced code blocks, inline code, URLs in link parentheses, and the table separator line pass through untouched.

2. Paste text, upload a file, or enter a page URL in the lower block. Editing starts immediately; there is no "Count" button.

3. In the URL check block, choose what to fetch. "Markup" is for checking the title, description, og data, and headings—in other words, the lines used in search and social media. "All visible text" is for processing the entire page: findings are then shown as text snippets around each edit, and the finished text can be fetched with a button or sent to the editing field.

4. Switch to "Changes". Each edit is marked, and the rule and previous version are visible under the cursor. If a rule fired in the wrong place, click the marker to undo that edit and recalculate the counters. Clicking the underlined location restores the edit.

5. Choose the output format. Layout developers need entities: you cannot distinguish a non-breaking space from a regular one by eye in finished text, while   is visible. Each of the three formats has copy buttons, and you can also download the result as a file.

6. "Apply to source" puts the result back into the left field, so you can continue editing it manually without moving text between panels. The bookmarklet in the last block edits a field directly in a third-party admin panel.

The formula and what it includes

• Quotes — A straight quote after a space, parenthesis, dash, or at the start of a line is treated as an opening quote; all others are closing quotes. The pairs for the first and second levels are selected from a list: guillemets, double quotation marks, English double quotes, English single quotes, and programmer-style quotes. The nesting counter survives tags, so «<b>"quote"</b>» does not receive two opening quotes in a row.
• Em dash — A hyphen surrounded by spaces, and a hyphen at the start of a line followed by a space, become an em dash. A hyphen inside a word is left alone: «из-за» and «кто-то» remain unchanged. In Markdown mode, a hyphen at the start of a line is never changed—it is a list marker there. In the German set, an en dash with spaces is used between words.
• En dash — A hyphen between two numbers becomes an en dash without surrounding spaces: 5-7 becomes 5–7. If a sequence contains multiple hyphens, as in a phone number or date, the rule does not apply.
• Non-breaking space — Added based on a word list, not length. Prepositions, conjunctions, particles, and any one-letter word are attached to the next word; «бы», «б», «ли», «ль», «же», and «ж» are attached to the previous word. Separately, a number is attached to the following word («10 кг», «25 января», «2024 год»), an abbreviation to a number («стр. 12»), a word to a number («дом 5»), an initial to a surname, № to a number, and «г.» to a city and a year. A space before a dash and before % ‰ ° ₽ is also made non-breaking. Short-word attachment is disabled for English and German because it is not customary there.
• Punctuation — Three dots become an ellipsis, (c) becomes ©, (r) becomes ®, (tm) becomes ™, the letter x between digits becomes the multiplication sign ×, and +- before a number becomes ±. The degree sign is attached to the number: «25 ° C» becomes «25 °C» with a non-breaking space. If a letter immediately follows a parenthesis, no replacement is made: «(С)ветлана» remains a word; (c) is also left alone in the list «(a), (b), (c)»; 0x1F is treated as a hexadecimal number.
• Math — <= becomes ≤, >= becomes ≥, != and <> become ≠, ~= becomes ≈, -> and --> become →, <- and <-- become ←, and +/- becomes ±. This is a separate rule that can be disabled; in text containing code fragments, it is better not to change arrows.
• Abbreviations and rubles — «т.д.», «т.п.», «т.е.», «т.к.», «т.ч.» receive a period, a non-breaking space, and the second letter. A year is attached to «г.» and «гг.». A number with «RUB» becomes a number with a non-breaking space and the ₽ sign. A non-breaking space is added after №.
• Phone numbers — A number beginning with +7 or 8, with parentheses, hyphens, or spaces, is normalized to one format: +7 (999) 123-45-67. The spaces inside the number are non-breaking, so it will not split across lines.
• Letter ё — Added according to a dictionary, and only where spelling with «е» is not a different word. «Все» and «всё», «небо» and «нёбо», «осел» and «осёл», and «берет» and «берёт» are not in the dictionary: those cases require meaning, not a list. This rule is disabled by default.
• Extra spaces — Consecutive spaces and tabs are collapsed into one; a space before , . ; : ! ? » … is removed; a space immediately after « „ ( is removed as well; and trailing spaces are stripped from lines. A space before a closing parenthesis is left alone: in Russian online messages, it is more often a smiley than a typing error. Indentation at the start of a line is left alone because it defines list nesting in Markdown. Markdown trailing spaces are not stripped either—two such spaces there mean a line break.

Example walkthrough: the line about the market

The input data is the same as the default field values: you can calculate it mentally and confirm that the tool is telling the truth.

• Text: ООО "Рога и копыта" - это 5-7 % рынка и не более того...
• Input: plain text; language detected as Russian
• All rules enabled except "Add ё"

1. Extra spaces: there are none here, so the counter stays at zero.

2. Punctuation: the three dots at the end collapse into a single ellipsis. One edit; the line becomes two characters shorter.

3. Quotes: the first follows a space and becomes an opening guillemet; the second becomes a closing guillemet. Two edits.

4. Dashes: the hyphen between spaces becomes an em dash; the hyphen between 5 and 7 becomes an en dash without spaces. Two edits.

5. Non-breaking spaces: after «и» inside quotes, according to the conjunction list; before the em dash; before the percent sign; after «и» and after «не» at the end. Five in total.

Output: ООО «Рога и копыта» — это 5–7 % рынка и не более того… In entity mode, the same line looks like ООО &laquo;Рога и&nbsp;копыта&raquo; &mdash; это 5&ndash;7&nbsp;% рынка и&nbsp;не&nbsp;более того&hellip;—which is the convenient form to put in a template.

Frequently asked questions

• Does the tool break HTML?

  No more. Switch the input to HTML, and the typographer will parse the markup before changing anything: it touches only text nodes, while tags, attribute values, existing entities, and the contents of script, style, pre, code, and textarea pass through unchanged. A quote inside class="…" stays straight. In "Plain text" mode, there is no parsing, and if the field contains tags, the tool will warn you directly in the summary.

• Why use entities when the actual characters are available?

  Because a non-breaking space in finished text looks like an ordinary space: you cannot copy it by eye from the result, and when passed through a third-party editor it may turn into a regular space. The &nbsp; notation is visible, survives any copy-paste and any templating engine. That is also why layout developers write &laquo; and &mdash;. The numeric form &#171; is needed where named entities are not supported, such as in XML.

• Why is a non-breaking space added after short words, but not after «он» and «мы»?

  Because it is now a list of words, not a length. The old rule counted letters and glued “он”, “мы”, and “их” just like “на” and “из”. Now prepositions, conjunctions, particles, and any one-letter word are attached to the next word, and the full list is shown directly below the settings—the rule must be verifiable.

• What happens to dates and phone numbers?

  The date 2024-01-15 stays as is: an en dash is used only when there is one hyphen between numbers. A phone number, on the other hand, is normalized to +7 (999) 123-45-67, with non-breaking spaces inside it so the number does not split across lines. This is a separate rule and can be disabled with a checkbox.

• How do I undo one edit instead of the entire rule?

  Switch the right panel to "Changes". Each edit is marked there; under the cursor, you can see which rule was triggered and what the text was before it. Clicking undoes that specific edit: the location becomes underlined, the counters are recalculated, and the rest of the text stays unchanged. Clicking the underlined text restores the edit. Undo states are cleared when you edit the source text; after that, the edit numbers are different.

• What exactly does the URL check inspect?

  Choose one of two options. "Markup" includes the title, description, og:title, og:description, and all H1–H6 headings: the dozen lines visible in search results and link previews. "All visible text" includes everything a person reads on the page: the server returns the complete text without scripts, styles, or hidden content. In both cases, the lines are processed by the enabled rules, and the table shows where straight quotes and hyphens instead of dashes occur and where a non-breaking space is missing. The findings are exported to CSV, while the page text can be placed in the editing field with one button or downloaded as a file.

• How do I read the findings from a full-text run?

  A whole paragraph will not fit in a table cell, so a piece of text is cut out around each edit—a couple of words on either side, at word boundaries. Adjacent edits go into one piece: “5-7 % of the market” is one finding with two edits, not two findings. The first column shows which rule was triggered, the two middle columns show the current and expected text, and the last shows how many edits are in the piece. The table shows the first 200 locations; the CSV contains them all.

• Why are the menu and footer included in the page text?

  Because the server returns everything visible, not just one article: it does not know which block you consider primary, and it will not guess—it would guess wrong and silently discard half the content. It is also more honest this way: menus and footers contain straight quotes and hyphens instead of dashes just as often as the main text, and no one usually fixes them. If you need only the article text, paste it into the field above.

• Is a bookmarklet safe?

  It is an ordinary bookmarklet with its code embedded: the full code is visible in the field below the button, so you can read it before dragging it to your bookmarks. It works only with the field containing the cursor, sends nothing anywhere, and makes no requests. Its rule set is limited to what fits in a bookmark URL: quotes, dashes, ellipsis, extra spaces, and non-breaking spaces.

• Is the text sent to the server?

  No. Editing, reading a file, and building the bookmarklet all happen in the browser. The only thing sent to the server is the page URL in the check block, and only after you click the button. When the page opens, it fetches its own markup, styles, scripts, and fonts once, all from the same domain.

• Can I share the result with a link?

  Yes. The settings—input, language, quote pair, rule set, and output format—are stored in the URL, along with the text itself if it is shorter than 600 characters. We do not put a long article in the link: such a URL would not survive any messenger. Use the "Download as file" button instead.