# CSV Numeric Column Extractor
Paste a CSV or TSV and get back only the columns that are fully numeric, as typed JSON arrays, records, CSV or a name list. Runs in your browser.
## Run it
- **CLI:** `gizza tool csv-numeric-column-extractor "id,name,score
1,Alice,9.5
2,Bob,7"`
- **Web:** https://gizza.ai/tools/csv-numeric-column-extractor/
- **Agents:** machine-readable descriptor (parameters JSON Schema) at https://gizza.ai/tools/csv-numeric-column-extractor/tool.json
## Inputs
- `data` — CSV or TSV text _(field)_
- `delimiter` — Delimiter _(field)_
- `header` — Header row _(field)_
- `output` — Output _(field)_
- `null_tokens` — Null tokens (comma-separated) _(field)_
- `allow_blanks` — Allow blank cells in a numeric column _(field)_
- `min_numeric_ratio` — Minimum numeric share _(field)_
- `normalize` — Accept $1,234.50 / 45% / (500) formatting _(field)_
## Output
- Numeric columns (text)
## Query parameters
Open the tool pre-filled and auto-run via URL:
- `data` — CSV or TSV text
- `delimiter` — Delimiter
- `header` — Header row
- `output` — Output
- `null_tokens` — Null tokens (comma-separated)
- `allow_blanks` — Allow blank cells in a numeric column
- `min_numeric_ratio` — Minimum numeric share
- `normalize` — Accept $1,234.50 / 45% / (500) formatting
Example: `https://gizza.ai/tools/csv-numeric-column-extractor/?data=id%2Cname%2Cscore%0A1%2CAlice%2C9.5%0A2%2CBob%2C7&delimiter=auto&header=auto&output=columns&null_tokens=NA%2CN%2FA%2CNULL%2Cnull%2CNone%2Cnan&allow_blanks=true&min_numeric_ratio=1&normalize=true`
---
## About this tool
Spreadsheets and database dumps mix identifiers, labels, dates and measurements in one table. This
tool reads a CSV or TSV and tells you which columns are **actually numeric** — then hands those
columns back as typed arrays you can drop straight into a chart, a statistics routine or a JSON
payload.
It sniffs the **delimiter** (comma, tab, semicolon or pipe), decides whether the **first row is a
header**, and tests every cell in every column. A column qualifies when all of its non-missing
values parse as numbers; loosen that with **Minimum numeric share** if a stray label shouldn't
disqualify an otherwise numeric column. Each qualifying column reports whether it is an `integer`
or a `float`, how many values it holds and how many were missing — and every rejected column comes
back in a `skipped` list with the reason and an offending example, so you never have to guess why a
column disappeared. Everything runs locally in your browser; the data is never uploaded.
### Worked example
Given this input:
```
id,name,score
1,Alice,9.5
2,Bob,7
```
the default **Typed arrays** output is:
```json
{
"delimiter": "comma",
"header": true,
"rows": 2,
"columns_total": 3,
"numeric_columns": 2,
"columns": [
{
"name": "id",
"index": 1,
"type": "integer",
"count": 2,
"missing": 0,
"numeric_ratio": 1,
"values": [1, 2]
},
{
"name": "score",
"index": 3,
"type": "float",
"count": 2,
"missing": 0,
"numeric_ratio": 1,
"values": [9.5, 7]
}
],
"skipped": [
{
"name": "name",
"index": 2,
"reason": "only 0 of 2 value(s) parse as numbers (0% < the 100% required)",
"example": "Alice",
"numeric_ratio": 0
}
]
}
```
Switch **Output** to *Column names only* for `id`, then `score` on the next line; to *Numeric
columns as CSV* for `id,score` / `1,9.5` / `2,7`; or to *Row objects* for
`[{ "id": 1, "score": 9.5 }, { "id": 2, "score": 7 }]`.
### Accounting formats
With **Accept $1,234.50 / 45% / (500) formatting** left on, exported figures still count as
numbers: thousands separators (`1,234.50` → `1234.5`), currency symbols `$ € £ ¥ ₹`, a trailing
percent sign (`45%` → `45`), parentheses negatives (`(500)` → `-500`) and trailing minus
(`250-` → `-250`). Turn it off when you want strictly plain numbers.
### Limits & edge cases
- Input is capped at **1 MB** of pasted text (about 1,000,000 characters). Split larger exports and
run the parts separately.
- Zero-padded codes such as `007` or a ZIP like `01234` are **not** treated as numbers, so
identifier columns don't get silently converted.
- `inf`, `Infinity` and `NaN` are treated as text, not numbers.
- Missing cells are the empty cell plus anything in **Null tokens** (matched exactly and
case-sensitively after trimming). They become `null` in the output and are excluded from the
numeric share.
- Ragged rows are padded: a row shorter than the widest row counts as missing in the trailing
columns. Wholly blank lines are dropped.
- Values are re-emitted as parsed numbers, so `1,234.50` comes back as `1234.5` and trailing zeros
are not preserved.
## FAQ
What makes a column count as "numeric"?
Every non-missing cell in it has to parse as a number. Set **Minimum numeric share** below `1` to
relax that — at `0.75`, a column where three of four values are numbers still qualifies, and the
value that didn't parse is emitted as `null`.
My column has a few blanks — will it still be extracted?
Yes. **Allow blank cells in a numeric column** is on by default, so gaps become `null` and the
column is still returned; the `missing` count tells you how many there were. Turn it off to require
a value in every row.
Why was my ID or ZIP column skipped?
Zero-padded values like `007` or `01234` are deliberately rejected: they are identifiers whose
leading zeros would be lost as numbers. The `skipped` list names the column and shows the value
that disqualified it.
How does it choose the delimiter and the header row?
With **Delimiter** on *Auto-detect* it parses the text once per candidate (comma, tab, semicolon,
pipe) and keeps whichever yields the most consistent column count, requiring at least two columns.
With **Header row** on *Auto-detect*, the first row is treated as a header unless one of its cells
is itself a number — so `10;20;30` on line one is read as data and the columns are named
`column_1`, `column_2`, `column_3`. Both can be forced.
Can I get the result as CSV instead of JSON?
Yes — set **Output** to *Numeric columns as CSV* for a comma-separated table containing only the
numeric columns (missing values become empty cells), or to *Column names only* for a plain list you
can paste into a query or a script.
Is my data uploaded anywhere?
No. The tool is compiled to WebAssembly and runs entirely in your browser tab — the CSV you paste
never leaves your machine.
## Related tools
- [Absolute value, sign, or negation for a whole column](https://gizza.ai/tools/absolute-value-transformer/): Paste a column of numbers and apply absolute value, sign extraction (-1/0/1), sign flipping, or force-negative to every value at once, with rounding and an audit table.
- [Adjacency Matrix Converter](https://gizza.ai/tools/adjacency-matrix-converter/): Convert a graph between edge list, adjacency matrix, and incidence matrix — directed or undirected, weighted or not. Free, private, runs in your browser.
- [Amazon Order Analyzer](https://gizza.ai/tools/amazon-order-analyzer/): Paste an Amazon order-history CSV export to summarize total spend by month, top items, and category breakdowns. Browser-only, private, with Markdown or JSON output.
- [ARFF Converter](https://gizza.ai/tools/arff-converter/): Convert Weka ARFF datasets to CSV and CSV tables back to ARFF locally — nominal attributes, numeric types, dates, sparse rows, missing values, and type rows.
- [Avro to JSON Converter](https://gizza.ai/tools/avro-to-json/): Decode Apache Avro Object Container Files (.avro / OCF) to JSON, NDJSON, or the embedded schema — no .avsc needed, free and private in your browser.