# NDJSON / JSON Lines Filter Filter and reshape NDJSON (JSON Lines) in your browser. Keep rows with a predicate, pick or rename fields, and export as JSONL, a JSON array, or CSV. ## Run it - **CLI:** `gizza tool ndjson-filter '{"name":"Alice","age":30,"city":"NYC","active":true} {"name":"Bob","age":25,"city":"LA","active":false} {"name":"Carol","age":40,"city":"NYC","active":true}'` - **Web:** https://gizza.ai/tools/ndjson-filter/ - **Agents:** machine-readable descriptor (parameters JSON Schema) at https://gizza.ai/tools/ndjson-filter/tool.json ## Inputs - `data` — NDJSON input (one JSON object per line) _(field)_ - `predicate` — Filter predicate (blank keeps every row) _(field)_ - `fields` — Fields to keep / rename (blank keeps whole records) _(field)_ - `format` — Output format _(field)_ - `invert` — Invert (keep non-matching rows) _(field)_ - `limit` — Max rows (0 = unlimited) _(field)_ - `skip_invalid` — Skip invalid JSON lines _(field)_ ## Output - Result (text) ## Query parameters Open the tool pre-filled and auto-run via URL: - `data` — NDJSON input (one JSON object per line) - `predicate` — Filter predicate (blank keeps every row) - `fields` — Fields to keep / rename (blank keeps whole records) - `format` — Output format - `invert` — Invert (keep non-matching rows) - `limit` — Max rows (0 = unlimited) - `skip_invalid` — Skip invalid JSON lines Example: `https://gizza.ai/tools/ndjson-filter/?data=%7B%22name%22%3A%22Alice%22%2C%22age%22%3A30%2C%22city%22%3A%22NYC%22%2C%22active%22%3Atrue%7D%0A%7B%22name%22%3A%22Bob%22%2C%22age%22%3A25%2C%22city%22%3A%22LA%22%2C%22active%22%3Afalse%7D%0A%7B%22name%22%3A%22Carol%22%2C%22age%22%3A40%2C%22city%22%3A%22NYC%22%2C%22active%22%3Atrue%7D&predicate=age%20%3E%2028%20and%20city%20%3D%3D%20NYC&fields=name%2C%20city%2C%20years%3Dage&format=ndjson&invert=true&limit=0&skip_invalid=true` --- ## About this tool NDJSON / JSON Lines Filter keeps just the rows and fields you want from newline-delimited JSON — one JSON object per line, the format that log pipelines, data exports, and streaming APIs emit. Write a small predicate to keep matching records, list the fields to keep or rename, and export the result as JSONL, a pretty JSON array, or CSV. Everything runs in your browser through WebAssembly — nothing is uploaded, so it is safe for private logs and exports. A predicate is one expression made of `path op value` clauses. Operators are `==`, `!=`, `>`, `>=`, `<`, `<=`, `contains`, `startswith`, `endswith`, and `~` (regex, also spelled `matches`). Join clauses with `and`, `or`, and `not`, and group them with parentheses. A bare path on its own (e.g. `active`) matches when that field exists and is truthy. Comparisons are numeric when both sides are numbers, otherwise string. Dotted paths (`user.name`) and array indexes (`items.0.id`) reach into nested data. Worked example — filter rows, then keep two fields: Input (NDJSON): ``` {"name":"Alice","age":30,"city":"NYC","active":true} {"name":"Bob","age":25,"city":"LA","active":false} {"name":"Carol","age":40,"city":"NYC","active":true} ``` Predicate: `city == NYC and age > 28` — Fields: `name, age` — Format: `ndjson` Output: ``` {"name":"Alice","age":30} {"name":"Carol","age":40} ``` Worked example — rename nested fields and export CSV: Input: ``` {"user":{"name":"Al","id":7},"city":"NYC"} {"user":{"name":"Bo","id":9},"city":"LA"} ``` Fields: `name=user.name, uid=user.id, city` — Format: `csv` Output: ``` name,uid,city Al,7,NYC Bo,9,LA ``` ## Limits and edge cases - Input is pasted text held in memory, so it is bound by your browser's available RAM — it is not built for streaming multi-gigabyte files. - Blank lines are always skipped. By default a line that is not valid JSON stops the run with a `line N: invalid JSON …` error; turn on **Skip invalid JSON lines** to silently drop malformed lines instead. - A missing path is treated as absent: `==` matches only `null`, `!=` matches any non-null, and the ordering/`contains` operators are false. In **Fields**, a missing path yields `null`. - `contains`, `startswith`, and `endswith` are case-sensitive substring tests; `~`/`matches` uses Rust `regex` syntax (no lookbehind/backreferences) and reports invalid patterns. - `and` binds tighter than `or`; use parentheses to force other grouping. `!` is not a negation operator — use the word `not` (or the **Invert** toggle for the whole predicate). - CSV columns are the first-seen union of the kept records' keys; cells containing a comma, quote, or newline are quoted RFC-4180 style, and `null` renders as an empty cell. Records that are not objects contribute empty cells. - **Max rows** stops filtering once that many rows are kept (`0` = unlimited); **Invert** keeps the rows that do NOT match the predicate. ## FAQ
What is NDJSON / JSON Lines, and how is it different from a JSON array? NDJSON (also called JSON Lines or JSONL) puts one complete JSON value — usually an object — on each line, with no enclosing brackets or commas between records. A JSON array wraps every element in `[ … ]` and separates them with commas. NDJSON streams well because each line is independent, which is why logs and exports use it. Set **Output format** to `array` to convert the kept rows back into a single pretty-printed JSON array, or `ndjson` to keep them line-delimited.
How do I filter on a nested or array field? Use a dotted path. `user.name == Al` reads the `name` key inside a `user` object, and numeric segments index into arrays, so `items.0.id > 100` looks at the first element of an `items` array. The same paths work in the **Fields** box, so you can flatten nested data — `id=user.id, city` pulls `user.id` up to a top-level `id` column. A path that does not exist in a record resolves to absent (missing) for the predicate and to `null` for field selection.
Can I use a regular expression? Yes. Use the `~` operator (or the word `matches`), for example `email ~ \.org$` to keep addresses ending in `.org`, or `name ~ ^A` for names starting with `A`. Patterns use the Rust `regex` engine, which supports the common syntax but not lookbehind or backreferences. An invalid pattern is reported as an `invalid regex` error rather than silently matching nothing.
How do I keep the rows that DON'T match? Turn on **Invert** to keep the complement of your predicate — every record that the predicate rejects. For example, predicate `active` with **Invert** on keeps only the inactive rows. Within a predicate you can also negate a single clause with `not`, e.g. `not (city == NYC)`; **Invert** flips the whole expression at once.
Is my data uploaded anywhere? No. The filtering runs entirely in your browser through WebAssembly — the text you paste never leaves your machine, so it is safe for private logs, database exports, and API payloads. There is no server round-trip and no sign-up.
## Related tools - [Data Format Converter](https://gizza.ai/tools/data-format-converter/): Convert data between CSV, TSV, JSON, and NDJSON (JSONL) in any direction — auto-detects the source, infers types, unions keys. Free, in-browser, no upload. - [Code Chunker](https://gizza.ai/tools/code-chunker/): Split Python, Rust, JavaScript, TypeScript, Go, Java, C/C++, C#, PHP, or Swift into line-ranged chunks that keep functions and classes intact. - [Code Formatter](https://gizza.ai/tools/code-formatter/): Beautify and re-indent minified or messy HTML, CSS, JavaScript, or JSON. Auto-detect the language, choose spaces or tabs, and format locally in your browser. - [CSS Autoprefixer](https://gizza.ai/tools/css-autoprefixer/): Free online CSS autoprefixer — paste CSS and add the -webkit-, -moz-, -ms-, -o- vendor prefixes browsers still need. Runs entirely in your browser. - [Directory Tree View](https://gizza.ai/tools/directory-tree-view/): Paste a du, find, or CSV file listing and get a clean indented tree with rolled-up folder sizes and file counts. Human, SI, or raw bytes. Free, in-browser.