Embedded JSON Field Extractor

How to use Embedded JSON Field Extractor

Recover selected properties from a JSON payload column without turning large integers into rounded JavaScript numbers. Each successful record retains the selected scalar type and original numeric lexeme; failures identify the data record and leave its appended fields empty under an explicit policy. The original CSV fields remain available as evidence.

1. Paste a rectangular CSV and choose the exact column containing one complete JSON value per record.
2. Map up to ten distinct new column headers to plain RFC 6901 JSON Pointers.
3. Choose stop-on-error or row annotation, then review typed provenance before downloading the formula-protected CSV.

Good to know

• At most 200,000 CSV characters, 1,000 records and 64 original columns. An embedded JSON cell is at most 20,000 characters; their combined source is at most 150,000. Parsing is bounded to 50,000 aggregate nodes, 40 levels and 10,000 nodes per JSON value.
• Mappings use at most ten new nonblank headers and plain JSON Pointers of at most 1,000 characters/40 tokens. ~0 and ~1 decode exactly; arrays require canonical nonnegative indices without leading zeros. URI-fragment pointers, absent values, containers and duplicate JSON object members are refused.
• Only selected string, boolean, null and number scalars are exported. Numbers retain their original JSON token spelling, including large integers and exponents. Null becomes an empty CSV cell with explicit null type evidence, so it remains distinguishable from an empty string in the report. No arithmetic or type coercion occurs.
• An errored record exports no mapped values, even if some pointers succeeded. stop refuses the whole result; annotate keeps original fields and adds _json_error. That header is reserved. CSV formula-prefix protection may add apostrophes; original source strings and scalar evidence remain unchanged in JSON.
• Processing is local. Output is limited to 1,000,000 text characters unless a stricter file limit is stated.