Re: reading and writing in bulk.

From: David Van Huffel <david.vanhuffel@teleatlas.com> > > >Does there exist a function in the C-api that can read and write to >informix tables in bulk (something like dbload) ? >This in order to increase performance for "dumping" and/or "database >filling" routines. For the moment, everything is working record per >record, but preformance is not really acceptable. We can work faster >writing it in ASCII and then afterwards loading it into Informix with >DBLoad. onpload (high performance loader) ______________________________________________________ Get Your Private, Free Email at http://www.hotmail.com

You can read in bulk using the ESQL/C FETCH ARRAY feature. This is documented and you can look at the code in my dbcopy.ec program for how to use it. Quickly you set the global variables FetArrSize (number of rows to fetch in a single FETCH operation) and FetBufSize (size of the transfer buffer used for array fetching, actually all fetches and puts), prepare and describe the statement, allocate an array of FetArrSize elements for each column returned, link the arrays into the sqlda structure, FETCH, and walk the column arrays in parallel to construct individual records. The two globals are mutually limiting, ie the number of rows returned is the lesser of FetArrSize or the number of rows that fit in FetBufSize (which must be <= 32767). See dbcopy.ec which is contained in the package utils2_ak in the IIUG Software Repository for an admittedly complex example of using the FETCH ARRAY feature. The closes thing to writing in bulk is to use an INSERT CURSOR and PUT to it after increasing the PUT buffer size using FetBufSize. The rows will actually be inserted when the buffer is flushed either when it fills or when you manually FLUSH it. The dbcopy.ec program also shows how this works which is pretty much how dbload does it. Using these two features dbcopy is 3x faster than with them disabled! As Obnoxio points out for single table loading you can use the onpload utility which can do low level and unlogged loads which are faster still. Art S. Kagel Obnoxio The Clown wrote: > > From: David Van Huffel <david.vanhuffel@teleatlas.com> > > > > > >Does there exist a function in the C-api that can read and write to > >informix tables in bulk (something like dbload) ? > >This in order to increase performance for "dumping" and/or "database > >filling" routines. For the moment, everything is working record per > >record, but preformance is not really acceptable. We can work faster > >writing it in ASCII and then afterwards loading it into Informix with > >DBLoad. > > onpload (high performance loader) > ______________________________________________________ > Get Your Private, Free Email at http://www.hotmail.com