Sort a CSV by any column, filter its rows on a condition, and transpose rows into columns — without opening a spreadsheet. The sort is type-aware, which matters more than it sounds: compared as text, 10 comes before 9, so a column of quantities sorts into an order nobody wants. Numbers are compared as numbers here, and text comparison folds case so Apple and apple sort together rather than capitals first. The header row stays at the top through every operation. Filters offer contains, equals, greater than, less than and empty checks, and the conditions apply before the sort so the two compose predictably. Drop a folder to process files in bulk and get them back as a ZIP; nothing is uploaded.
| Condition | Matches when the cell |
|---|---|
| contains | Includes your text, ignoring case |
| equals | Is exactly your text, ignoring case |
| does not equal | Is anything else |
| greater than | Sorts after your value — numerically if both are numbers |
| less than | Sorts before your value |
| is empty | Is blank or only whitespace |
| is not empty | Has any content |
No. Sorting and filtering happen in your browser, and nothing is transmitted.
Because they compare as text, where 10 precedes 9 — the character 1 comes before the character 9. Each cell is inspected here first, so a numeric column sorts numerically. A column mixing numbers and words keeps the numbers together rather than scattering them through the text.
No. It is held in place through sorting, filtering and transposing. Sorting the whole array and letting the header land in the middle is the classic version of this bug.
Filter, then sort, then transpose. The order matters: filtering after a transpose would filter by row while the control still says column, which would contradict what the interface told you.
They are padded rather than dropped, so transposing does not silently lose data from a short row.
No. Contains and equals both ignore case, which is almost always what you want when filtering exported data by hand.
Sorting is type-aware, which is the difference between a useful sort and a confusing one: compared as text, 10 sorts before 9 because the character 1 precedes the character 9, and a column of quantities comes out in an order nobody asked for. Each cell is inspected first, and when both sides read as numbers they are compared numerically. A column mixing the two keeps all its numbers on one side rather than interleaving them, because an interleaved result looks like a bug whichever way you read it. Text comparison uses locale-aware collation with case folding, so Apple and apple sort together instead of all capitals coming first. The header row is held in place through every operation, which the naive implementation gets wrong the moment it sorts the whole array and the header lands in the middle of the data. The operations apply in a fixed order — filter, then sort, then transpose — because filtering after a transpose would filter by row while the control still says column, and the result would contradict the interface. Transposing pads ragged rows rather than dropping the short ones, so nothing is silently lost when a row has fewer cells than the header. Parsing is full RFC 4180, so quoted fields containing the delimiter survive; sorting by a column is meaningless if the parse has already put values under the wrong header. Everything runs in your browser and no file is uploaded.