Download a reference table's items as a CSV. The mirror of /csv-upload.
Returns the file itself -- text/csv, as an attachment named after the table's
slug, so a browser saves browsers.csv -- rather than a JSON envelope. Nothing but
the items is in it: title, description, visibility and history are not part of the
upload format and so have no column here.
The format is the one POST /reference-tables/{id}/csv-upload documents, and the
output is valid input to it: the header names the fields in stored order, each
following line is one item, and a field an item does not carry is a blank cell.
Exporting, editing in a spreadsheet and uploading the result is the intended loop.
One item row may export as several lines. A row can hold more than one value for
the same field -- the canonical browsers list is a single row naming three of them --
and a line has one cell per field, so those values are written down the column. Read
back, that is the same table: expansion collects each column's values across every row
and ANDs the columns, so the distribution of values between rows is not something the
format carries. Row-for-row identity is what is lost; the predicate is not.
Because the columns are ANDed, this file is not a grid of tuples, and editing it
as one is the mistake to watch for. Adding a value to a second column constrains every
match, not just the line it is on.
A table with no items exports an empty file, with a 200 rather than an error --
there is nothing wrong with asking. Uploading it back is a 400: emptying a list stops
every rule that uses it, so it has to be asked for explicitly through PATCH.
Readable tables only, which includes published Nebulock-curated ones -- taking a
curated list away to work on is the point of sharing them. Anything else is a 404.
| Time | Status | User Agent | |
|---|---|---|---|
Retrieving recent requests… | |||