Regex to JSON
Paste lines of text, write a regex with named capture groups, and get one JSON object per line — group names become keys. Coerce numbers and booleans, keep or drop unmatched lines, and output a JSON array or NDJSON. Runs in your browser; nothing is uploaded.
About this tool
Regex to JSON turns unstructured lines — log files, command output, exports — into structured JSON. Write a regular expression with named capture groups and each line becomes one JSON object whose keys are the group names, in the order they appear in the pattern.
- Named groups are the schema:
(?<name>…)(or Python-style(?P<name>…)) defines the JSON key; numbered groups are ignored. A pattern with no named groups is rejected with a hint. - Unmatched lines are yours to decide: Skip them (default), Keep
them as
{"_raw": "<line>"}so no data is lost, or Fail with an error naming the first non-matching line — handy for validating a format. - All matches per line emits one object per match instead of one per line —
ideal for
key=valuepairs repeated on a single line. - Coerce types converts captures that look like plain numbers (
42,-3.14),true/false, ornullinto real JSON types. Values with leading zeros (007) or scientific notation stay strings so IDs are never mangled. - Output: a pretty JSON array, a compact single-line array, or NDJSON (one JSON object per line — JSON Lines) for piping into other tools.
Worked example
Input lines:
10.0.0.5 GET /index.html 200 512
10.0.0.9 POST /api/login 401 87
Pattern (with Coerce types on):
(?<ip>\S+) (?<method>[A-Z]+) (?<path>\S+) (?<status>\d{3}) (?<bytes>\d+)
Output:
[
{ "ip": "10.0.0.5", "method": "GET", "path": "/index.html", "status": 200, "bytes": 512 },
{ "ip": "10.0.0.9", "method": "POST", "path": "/api/login", "status": 401, "bytes": 87 }
]
Limits and edge cases
- Input is capped at 1 MB (1,000,000 bytes); larger text is rejected with an error.
- The syntax is the Rust
regexflavour — linear-time matching with no catastrophic backtracking, but no look-around ((?=…),(?<=…)) and no backreferences (\1). - Blank and whitespace-only lines are ignored entirely — they are neither parsed nor counted as unmatched. Windows (CRLF) line endings are handled.
- A named group that did not participate in a match (an optional group) is
emitted as
null, so every record keeps the same keys. - If nothing matches, the result is an empty array
[](empty output for NDJSON) — not an error, unless Unmatched lines is set to Fail.
Everything runs locally in your browser via WebAssembly — neither your text nor your pattern is uploaded.
FAQ
Why do I get "the pattern has no named capture groups"?
This tool builds JSON keys from group names, so at least one group must be
named: write (?<status>\d{3}) instead of (\d{3}). Both the (?<name>…)
and Python's (?P<name>…) spellings work. Plain numbered groups are ignored —
if you only want a flat list of matches, the separate regex-extract tool does
that.
What happens to lines that don't match the pattern?
By default they are skipped. Choose Keep as {"_raw": line} to carry them
through the output unparsed (so you can spot format drift downstream), or
Fail with an error to stop at the first non-matching line — the error names
the 1-based line number and shows the line, which makes it a quick format
validator. Blank lines never count as unmatched.
How does type coercion decide what becomes a number?
With Coerce types on, a capture becomes a JSON number only if it is a plain
integer (42, -7) or plain decimal (3.14); exact true/false become
booleans and exact null becomes null. Deliberately conservative: leading-zero
values (007), scientific notation (1e5), thousands separators (1,000),
and integers too big for 64 bits all stay strings, so identifiers survive
untouched. Off by default — everything is a string.
Can one line produce more than one JSON object?
Yes — tick All matches per line. Instead of taking only the first match,
every match on a line emits its own object. With the pattern
(?<key>\w+)=(?<value>\S+) the line host=web1 region=eu yields two objects.
The order is left-to-right within each line, top-to-bottom across lines.
Why doesn't my lookbehind or backreference work?
The engine is the Rust regex flavour, which guarantees linear-time matching —
a hostile pattern can never hang the page. The trade-off is that look-around
and backreferences are unsupported and rejected with a syntax error. In
practice a named group around the part you want, plus anchors (^, $, which
match per line here since each line is matched on its own), covers most
patterns.
Developer & Automation Access
Run it from the terminal
Same engine as this page, headless — via the gizza CLI:
gizza tool regex-to-json "10.0.0.5 GET /index.html 200 512
10.0.0.9 POST /api/login 401 87" 'pattern=(?<ip>\S+) (?<method>[A-Z]+) (?<path>\S+) (?<status>\d{3}) (?<bytes>\d+)'New to the CLI? Get gizza →
Open it by URL
Pre-fill and auto-run this tool with query parameters — the names match the API/CLI:
https://gizza.ai/tools/regex-to-json/?text=10.0.0.5%20GET%20%2Findex.html%20200%20512%0A10.0.0.9%20POST%20%2Fapi%2Flogin%20401%2087&pattern=%28%3F%3Cip%3E%5CS%2B%29%20%28%3F%3Cmethod%3E%5BA-Z%5D%2B%29%20%28%3F%3Cpath%3E%5CS%2B%29%20%28%3F%3Cstatus%3E%5Cd%7B3%7D%29%20%28%3F%3Cbytes%3E%5Cd%2B%29&ignore_case=true&all_matches=true&unmatched=skip&coerce_types=true&output=jsonMachine-readable descriptor: tool.json — title + parameters JSON Schema for agents.