Numeric Row Deduplicator
Remove duplicate rows from a numeric table, comparing each cell by its numeric value — so 1, 1.0, 1.00, +1 and 1e0 all collapse to one row, unlike a plain text deduper. Key on chosen columns, round away float noise, and keep the first or last occurrence. Runs entirely in your browser; nothing is uploaded.
About this tool
The Numeric Row Deduplicator removes duplicate rows from a numeric table (CSV or TSV), but
unlike an ordinary text deduplicator it compares each cell by its numeric value. That means
1, 1.0, 1.00, +1, 1e0 and 100e-2 are all recognised as the same number, so rows that
only look different collapse into one. A plain string-based deduper would keep every one of them.
Paste your rows, optionally pick the columns to key on, and the tool keeps the first occurrence of each numeric row (in its original order) and drops the rest. Everything runs locally in your browser — nothing is uploaded.
Worked example
Input (one row per line):
1,2
1.0,2.0
1.00,2
3,4
Output (De-duplicated rows):
1,2
3,4
Rows 1–3 are the same pair of numbers written three ways, so only the first survives; the 3,4
row is unique and is kept.
What you can control
- Key columns — match on a subset of columns by 1-based number (
1,3) or, with a header, by name (id,region). Blank keys on the whole row. - Header — when on, the first row is kept verbatim, excluded from the scan, and its names can be used as key columns.
- Delimiter —
,(default),tab,;(semicolon) or|(pipe), or any single character. - Rounding tolerance — round every numeric cell to N decimals before comparing so float-noise
near-duplicates collapse (
0.30000000000000004and0.3are equal at 2 decimals).-1compares the exact value. - Keep first / last — keep the earliest or the latest occurrence of each duplicate; either way the surviving rows stay in their original order.
Limits & behaviour
- Comparison is by value: two cells are equal when they parse to the same
f64(after any rounding). Non-numeric cells (text, currency symbols, blanks) fall back to a trimmed-string compare, so mixed tables still de-duplicate sensibly. - Rounding tolerance accepts 0–12 decimal places (or
-1for exact). - Very large numbers beyond double-precision range and special tokens like
NaN/Infare treated as text, not numbers. - This is exact numeric equality, not fuzzy matching — for approximate string matching use a fuzzy deduplicator instead.
FAQ
How is this different from a normal CSV duplicate remover?
A normal remover compares rows as raw text, so 1.0 and 1.00 are two different rows and both
survive. This tool parses each cell as a number and compares the values, so every textual form
of the same number collapses to a single row. That is the whole point of a numeric deduplicator.
Can I de-duplicate on just some columns?
Yes. Put the columns you want to key on in Key columns — 1-based numbers like 1,3, or header
names like id,region when Header is on. Rows that match on those columns are treated as
duplicates even if other columns differ, and the first (or last) matching row is kept.
What does the rounding tolerance do?
Set it to a number of decimal places and every numeric cell is rounded to that precision before
comparing. This collapses near-duplicates caused by floating-point noise — for example
0.30000000000000004 and 0.3 become equal at 2 decimals. Leave it at -1 to compare the exact
value with no rounding.
What happens to non-numeric cells?
Cells that do not parse as a number (labels, currency symbols, blanks, NaN) are compared as
trimmed text instead. So a table mixing a region column with numeric measurements still
de-duplicates correctly — the text columns match by text, the number columns match by value.
Is my data uploaded anywhere?
No. The tool is compiled to WebAssembly and runs entirely inside your browser tab. Your rows never leave your device, which makes it safe for sensitive or proprietary data.
Developer & Automation Access
Run it from the terminal
Same engine as this page, headless — via the gizza CLI:
gizza tool numeric-row-deduplicator "1,2
1.0,2.0
1.00,2
3,4"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/numeric-row-deduplicator/?data=1%2C2%0A1.0%2C2.0%0A1.00%2C2%0A3%2C4&columns=1%2C%203&header=true&delimiter=%2C&precision=-1&keep=firstMachine-readable descriptor: tool.json — title + parameters JSON Schema for agents.