Drop a CSV in and clean it: trim the spaces around cells, drop blank rows, remove duplicate rows, throw away columns that are empty everywhere, and normalise the headers into snake_case. The row count before and after is shown so you can see what was actually removed rather than trusting it. The ordering of these steps is not arbitrary — cells are trimmed before anything is judged empty or duplicate, because otherwise a row of three spaces survives 'remove blank rows' and 'Ada' and 'Ada ' count as two different people. Whole folders can be dropped at once and downloaded back as a ZIP. Files are read on your device and never uploaded, which matters given that the CSVs most in need of cleaning are customer and contact lists.
| Step | Why it runs there |
|---|---|
| Collapse whitespace | Before anything measures the cell |
| Trim cells | So a row of spaces reads as blank |
| Drop blank rows | After trimming, or they survive |
| Drop blank columns | Header counts, so named columns stay |
| Drop duplicates | After trimming, or 'Ada ' looks unique |
| Normalise headers | Last, on the row that survived |
No. It is parsed and cleaned in your browser, and nothing is sent anywhere. That is the reason to use this rather than pasting a customer list into a server-side service.
No. Parsing follows RFC 4180, so a quoted field containing commas, doubled quotes or a line break inside it survives a round trip intact. Quoting is recalculated on output, so a field only carries quotes if it still needs them.
Because cells are trimmed before emptiness is judged, which is almost always what you want. A row containing nothing but whitespace is blank to a person, and leaving it in defeats the purpose of the option. Turn off 'trim spaces' if you need whitespace preserved exactly.
The first. Later identical rows are dropped, so the original order of the rows you keep is unchanged. The header row is never counted as a duplicate of a data row.
One where every cell is blank, including the header. A column with a header but no data yet is kept, because dropping it would change the shape of a file that is still being filled in.
A byte-order mark is an invisible marker some tools, notably Excel, write at the start of a UTF-8 file. It becomes part of the first header, which is why a first column sometimes refuses to match by name. It is removed automatically here.
The cleaning steps run in a fixed order, and the order is the part that decides whether the result is right. Cells are trimmed before emptiness is judged, because a row containing three spaces is a blank row to a human and is not one to a naive check — clean in the other order and it survives. Duplicates are removed after trimming for the same reason: without it, Ada and 'Ada ' count as two different people, which is exactly the duplicate you were trying to find. The header row is never treated as a duplicate of itself, so a single-column file whose header repeats a value does not quietly lose its header. A column counts as empty only when every cell including the header is blank, so a named column that has no data yet is kept — dropping it would change the shape of a file someone is still filling in. Parsing is full RFC 4180: quoted fields containing the delimiter, doubled quotes, embedded newlines and CRLF all survive a round trip, and a UTF-8 BOM is removed rather than becoming part of the first header, which is the usual reason a first column mysteriously will not match by name. Header normalisation lowercases and replaces every run of non-alphanumeric characters with a single underscore, trimming any that end up at the edges, which is what makes a header usable as a database column name. Everything runs in your browser — the file is never uploaded, which matters because the CSVs people need to clean are usually customer lists.