> For AI agents: the complete documentation index is available at /en/llms.txt, the full documentation bundle is available at /en/llms-full.txt. *** title: Multi-Function Text Processor head: - - meta - property: og:title content: Multi-Function Text Processor - Regex matching, filtering, sorting, batch formatting | Tools By AI description: An all-in-one text toolbox. Regex match, delete lines, dedupe, sort, merge, batch prefix/suffix, letter case, JSON beautify and smart splitting can be chained into a step pipeline and saved as presets. Runs entirely in your browser — nothing is uploaded. faq: - q: What operations does this tool offer? a: 18 of them: regex match / remove / replace, delete lines by keyword, deduplicate whole lines, sort ascending or descending, reverse order, swap adjacent lines, merge into one line, add prefix/suffix, letter case (UPPERCASE / lowercase / Title case / Sentence case), format text, smart split, tidy CLI text, JSON beautify, pair extract, extract links with metrics, and the Weibo link converter. Operations chain: each click appends a step, and the previous step's output feeds the next one. - q: What if I can't write regular expressions? a: The REGEX band has four preset chips — URL (no params), URL (loose match), Remove index (. 、), Extract JSON keys. One click fills the regex field. For anything harder, hand the requirement plus a sample and the expected output to an AI, then paste the result back to test it here. - q: What does the Smart trim switch actually do? a: While on (the default), almost every operation first strips leading/trailing whitespace from each line and skips blank lines. Turn it off — the switch then reads Raw mode — and the original line structure is kept. The one operation that depends on blank lines is Tidy CLI text, and it ignores this switch. - q: Difference between "Remove matches" and "Run match"? a: Run match pulls every regex hit out and lists them line by line; Remove matches does the opposite — deletes the hits and keeps the rest, collapsing runs of blank lines down to two. - q: How is the step chain different from "Result ➔ Source"? a: The chain already is a pipeline: one click per step, evaluated top to bottom, each step's output feeding the next — no manual round-trips. "Result ➔ Source" is for when you want the current result to become the new input: it writes the result back into Source and clears the steps. - q: Can presets be exported or imported? a: No. A preset is a saved step chain living in your browser's localStorage, not a file. Settings import/export does not cover them, so on another machine you rebuild the chain — or note down its steps. - q: Can I get the processed text out? a: The Result panel has Copy and Export (saved as `text-processed.txt`), plus Result ➔ Source to feed the output back in. Everything runs locally; nothing is uploaded. Past roughly 1,000,000 characters the input box flips to read-only with a note, while processing and export still cover the full text. appUrl: [https://tools.newzone.top/en/text-toolbox](https://tools.newzone.top/en/text-toolbox) appName: Multi-Function Text Processor image: [https://docs.newzone.top/img/text-toolbox-en.webp](https://docs.newzone.top/img/text-toolbox-en.webp) howto: name: Cleaning text with the Multi-Function Text Processor steps: - name: Paste or upload your text text: Put it in the Source box, or use "Click or drag file here to upload" to load a file. - name: Click operations to build a step chain text: Pick operations from the LINES / CASE / REGEX / FORMAT bands in the order you need — each click appends a step and the result recomputes live. - name: Tune a single step text: Click that step's chip above the result and edit its regex, keywords, separator and so on; only that step changes. - name: Reuse or export text: Save a handy chain with "Save as preset"; hand the output over with Copy or Export as a .txt. *** # Multi-Function Text Processor [Multi-Function Text Processor](https://tools.newzone.top/en/text-toolbox) gathers the text cleaning, extraction and formatting you do most often onto one page. The point isn't a wall of separate buttons — it's an **operation catalog plus a step chain**: every click appends a step, the chain runs top to bottom, and each step's output is the next step's input, recomputed as you go. ![Multi-Function Text Processor interface](/img/text-toolbox-en.webp "Multi-Function Text Processor in use") ## What it's for - **Pull specific content out**: URLs, JSON keys, fields matching a pattern - **Strip the noise**: ad lines, quote residue, HTML tags, blank lines, duplicate lines - **Rescue text copied from a terminal**: remove the leading bar (`▎`) and hard wraps from Claude Code / terminal output and get paragraphs back - **Batch formatting**: prefix/suffix every line into Markdown lists, CSV, SQL `IN (...)` - **Normalize case**: UPPERCASE / lowercase / Title case / Sentence case - **Reorder data**: sort ascending or descending, reverse, swap adjacent lines - **Collapse lines into one**: join a column with any separator for pasting into a sheet or query - **Multi-step jobs**: filter, then regex, then affixes — click them in order and the chain runs the whole thing ## Layout Two columns. Left column, top to bottom: 1. **Source** panel — input, upload, the Smart trim switch, Clear all 2. **Result** panel — read-only output, with the chain's step chips just above it 3. **PRESET** — Select a preset, Save as preset, Update preset, Delete preset 4. **LINES** (_Line by line: sort · delete · dedupe · merge · affix_) — Sort / Duplicates / Delete / Merge / Add prefix/suffix 5. **CASE** (_Letter case_) — the four case buttons Right rail: - **REGEX** (_Regex engine_) — regex field, Presets chips, Match mode, Replace, Run match / Remove matches / Extract links with metrics, Pair extract - **FORMAT** (_Convert_) — Smart split, Format text, Tidy CLI text, JSON beautify, Weibo link converter Below 960px it collapses to a single column. The high-frequency line and case operations sit nearest the result; the more specialized regex and conversion work lives in the rail so it can't push the result off screen. ## The Source panel - **Paste**: drop text straight into the Source box (placeholder "Or type/paste text here") - **Upload**: the "Click or drag file here to upload" control in the panel header accepts TXT, MD, JSON, CSV and other text files; the loaded name shows as a chip you can remove on its own - **Smart trim / Raw mode**: the header switch. On (default) most operations first trim each line's surrounding whitespace and skip blank lines; off it reads **Raw mode** and keeps the original structure - **Clear all**: clears the input and uploaded file, and clears the step chain too - The footer shows live "characters · lines" counts Past roughly 1,000,000 characters the input box turns read-only with an explanation, while processing and export still work on the full content. ## The step chain The chain is the backbone — understand it and you never have to memorize button order: - **A step is added by clicking its operation.** At that moment the console fields' current values are **snapshotted** into the step; editing the fields afterwards does not silently rewrite steps you already added. - **Execution order is click order**, top to bottom. Evaluation starts from the source text and each step's output feeds the next, so five clicks aren't five manual runs. - **Async steps are awaited in sequence**: Smart split and the Weibo link converter take time, and the chain waits for them before continuing. - **Source text is never modified**: every operation reads the source and writes to the result. - The chain is visible as a row of chips above the result — `{count} steps applied` on the left, **Clear steps** on the right. Each chip reads `operation · parameter summary`, e.g. `Delete · ad,promo`, `Add prefix/suffix · 1000----`. - **Clicking a chip opens its parameter popover**, and the edit applies to that step alone; the `×` on the chip is "remove this step". - **There is no drag-to-reorder.** To change the order, remove the step and click again in the order you want — or use Result ➔ Source to restart from the current output. - **A step that fails validation never enters the chain**: clicking "Delete these lines" with an empty keyword field, or with an invalid/empty regex, only shows a warning. - The chain persists in localStorage under `text-toolbox-chain` and survives a reload. On restore, unrecognized operations are dropped and missing parameters fall back to factory defaults. :::tip When to use Result ➔ Source The chain is re-editable: tune a step's parameters, delete a step, add one. Reach for **Result ➔ Source** only when the current output should become the **new input** for a different run — it writes the result back into Source and clears the steps. ::: ## PRESET — presets A preset is a **saved step chain** including each step's own parameters. - **Select a preset**: every dropdown entry shows its step sequence on a second line (longer chains fold to `… +N`) - **Save as preset**: enabled when the chain isn't empty; type a **Preset name** in the dialog - **Update preset / Delete preset**: only for presets you saved yourself - Loading a preset **replaces** the current chain rather than appending - If your chain no longer matches the loaded preset, the dropdown reads "Modified (was: name)" as a reminder that you haven't saved it back - Three built-ins ship with the page and cannot be edited or deleted: - **Extract links, reversed** — Run match (loose URL) › Reverse order › Merge into one line (separator `,`) - **Extract URLs, dedupe, sort** — Run match (strict URL) › Deduplicate › Sort ascending - **HuggingFace to ModelScope** — Replace: `https://huggingface.co` → `https://modelscope.cn/models` - Your own presets and the active selection live in localStorage keys `text-toolbox-presets` and `text-toolbox-activePresetId`. **Presets can't be imported or exported** — on another browser you rebuild the chain :::warning Think before editing a chain Built-in presets are read-only so someone else's chain can't quietly drift in your hands. Want a variant? Load it, then save under a new name instead of overwriting. ::: ## LINES — line by line The band header names what's inside: _sort · delete · dedupe · merge · affix_. Each row puts parameters on the left and the verb on the right, so a row reads as "with this condition → do this". ### Sort - **Sort ascending / Sort descending**: Unicode string order. The button label shows the direction the _next_ click will apply, so it flips as you click - **Reverse order**: flip the line order - **Swap adjacent lines**: exchange lines in pairs. Blank lines are dropped first (the remaining count must be even, otherwise it reports an error), then a blank line is re-inserted between each pair. Useful when a metric sits on the line above its link and you plan to follow with "Extract links with metrics" ### Delete One condition field feeds one button: **Delete · \[match mode] · \[content] → Delete these lines**. - **Match mode** (segmented control): - **Contains**: delete a line if any keyword appears in it (substring) - **Whole line**: delete only exact line matches (ignoring surrounding whitespace) - **Content to delete**: separate several conditions with **commas or line breaks** (e.g. `ad,promo,group id`); stored locally - **Delete these lines**: removes the hits, keeps the rest, writes to the Result panel, leaves Source alone ### Duplicates - **Deduplicate** keeps the first of each identical line and preserves blank lines - It does **not** read the Delete row's keyword field — hence its own row with no input ### Merge - **Merge into one line** joins all lines into one - **Separator** (same row): what to join with; **leave it empty for direct concatenation**, and `\t`, `\n`, `\s`, `\\` escapes are supported - Example: `, ` turns a column into `A, B, C`; `\t` gives a tab-separated line you can paste into one Excel row ### Add prefix/suffix - **Prefix content**: prepended to every line; factory default is `1000----` (and the field can be cleared) - **Suffix content**: appended to every line; factory default is empty - The button on this row is labelled **Apply** — "Add prefix/suffix" is the row label - Example: prefix `- `, empty suffix → plain text becomes a Markdown list - Example: prefix `'`, suffix `',` → a string list becomes a SQL `IN (...)` clause ## CASE — letter case One operation, four quick buttons, applied line by line without touching blank lines or indentation: - **UPPERCASE** / **lowercase** - **Title case**: capitalize each word - **Sentence case**: capitalize the first letter of each sentence, lowercase the rest ## REGEX — regex engine Controls top to bottom: - **Regex field** (placeholder "Enter regular expression..."): any JavaScript regex; the factory value matches strict URLs - **Presets**: four chips that fill the field and set the flags each pattern needs - **URL (no params)** — plain `https://` links, no trailing punctuation - **URL (loose match)** — also handles parentheses, semicolons and friends - **Remove index (. 、)** — leading `1. `, `2、`, `3) ` - **Extract JSON keys** — every key name, in multiline mode - **Match mode**: global (g) / multiline (m) / case-insensitive (i); only `g` is on by default - **Replace**: a "Replace with…" field plus the **Replace** button. Replacement text is a **literal** — no `$1` group references - **Run match**: lists every hit on its own line - **Remove matches**: deletes hits, keeps everything else, collapses runs of blank lines to two - **Extract links with metrics**: groups by seven **Chinese** metric words (点赞 / 转发 / 评论 / 播放 / 差 / 曝光 / 阅读) and outputs `link,value` per line — so it only works on text containing those Chinese words - **Pair extract**: the **Extract & pair** button with "Pairing regex" and "Template" fields. The anchor is the shared regex field above (there is no separate anchor input any more): it takes the first match per line, the pairing regex then captures outside the anchor on the same line (falling back to the next line), and output follows the template where `$0` is the anchor and `$1` the captured pair. When every paired value is numeric, a total is appended in the result ## FORMAT — convert - **Smart split**: reflows long text into paragraphs, with **Chinese paragraphs / English paragraphs** on the adjacent segmented control (English uses the `compromise` library for sentence boundaries; Chinese follows paragraph rules) - **Format text**: removes blank lines, and trims each line's ends according to the Smart trim switch - **Tidy CLI text**: purpose-built for text copied out of Claude Code / terminals — strips the leading quote bar (`▎ ▌ │`) and indentation and rejoins hard-wrapped lines (Chinese joins directly, English gets a space), keeping list items and code blocks line by line. **It is the only operation that depends on blank lines**, so if an earlier step already dropped them the click warns you on the spot ("previous step dropped blank lines") and the step is skipped - **JSON beautify**: lenient parsing (unquoted keys, single quotes, comments) plus 2-space indentation - **Weibo link converter**: turns mobile Weibo (`m.weibo.cn`) links into PC `weibo.com` links and appends the blogger's profile URL. Links carrying only a uid resolve locally; detail pages need a network lookup for uid/bid, and entries that fail to resolve are labelled in the output text with the reason (no uid, uid/bid not found, network failure, unrecognized link) :::tip The Weibo converter needs the network Resolving Weibo detail pages sends requests, so it is slow and the chain waits for it. Offline or behind a blocked network this step fails while the rest of the chain still works. Smart split is local — it is async because it has to scan sentence boundaries, not because it calls out. ::: ## The Result panel - Titled **Result**, permanently read-only (a lock plus "Read-only mode" explains it). Change the source or a step rather than trying to edit output in place - With an empty chain, wide screens show a placeholder: "Run an action below — the result appears here" - Three actions, left to right: **Result ➔ Source** (write output back and clear the steps), **Copy**, **Export** (downloads `text-processed.txt`) - The footer shows the same "characters · lines" counts ## Working tips - **Plan the chain backwards from the goal**: "clean then affix" is Delete → Deduplicate → Add prefix/suffix, clicked once each in that order - Save a chain you use often as a preset; the three built-ins are ready to run - Test on a small sample first and read each chip's parameter summary to confirm it matches your intent - Can't write the regex? Give an AI the requirement, a sample input and the expected output - Prefer chaining over repeated "Result ➔ Source" round-trips — the latter clears your steps Everything runs in your browser and **nothing is uploaded** (the Weibo link converter is the one step that makes network requests), so sensitive text is safe to process here.