Simon Slavin wrote: > On 11 May 2012, at 3:36pm, Scott Ferrett > <sc...@ferrettconsulting.com> wrote: > >> If this is not possible, I can restrict this bit of code to only work on >> UPDATE statements. But that still leaves me with the problem of needing >> the rowid of the row being updated. > > The only supplied function is the one which returns the rowid from the most > recent INSERT. There are no equivalents for UPDATE or SELECT. Short of > parsing the SQL statements I don't know of a way to figure it out. > > I do have one application which edits SELECT statements as follows: after the > 'SELECT ' at the beginning, insert 'rowid,'.
SELECT * FROM foo ORDER BY 1; -- oops, adding column breaks ORDER BY SELECT * FROM foo JOIN bar; -- oops, "rowid" ambiguous without "foo." or "bar." SELECT * FROM foo GROUP BY t; -- oops, "rowid" is from random row SELECT * FROM foo UNION SELECT * FROM bar; -- oops IMO, such SQL statement manipulation is way too fragile to be used in any semi-general purpose library. You'd better just export low-level blob API and let library user worry about getting valid row-id, blob handle invalidation rules and all. > When it gets the results it > strips the first column off of the results from the SELECT before returning > the other columns to the calling function. This works perfectly in the > context of that app, which never uses JOIN, knows that every table has a > "rowid" column, and knows that the calling function will never care about > values in the "rowid" column. > > Simon. _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
