Paul J Stevens <[EMAIL PROTECTED]> said: > Mikhail Ramendik wrote: [snip] >> This is the only part that, in case of a header search, uses the data >> from more than just the header. What is this for? Why do we have to >> search "children"? [snip] > Shouldn't we restrict this looping over subparts to TEXT or BODY type > searches. In that case, with header-searches we can safely skip looping > over subparts.
I agree 100%. It would never occur to me to use a header search to find, for example, a message that was included in a mailing list digest. Also, I was thinking about the REGEXP idea with respect to parsing only a subset of headers. REGEXP isn't universally support; we would need to do a hack similar to the date function hack, where the db.c uses an external symbol that is defined by the database shim as needed for that database. It works well, but I wonder if it wouldn't be easier to just do this: SELECT * FROM messageblks WHERE messageblk LIKE '%From:[EMAIL PROTECTED]'; Since LIKE works everywhere, and doesn't fire up a regex engine... well, benchmarking is in order, I think ;-) Aaron --
