Re: return # of rows

Manuel Lemos wrote: > > Hello Steven, > > >Why not do a $query = "SELECT COUNT(ROWNAME) FROM TABLENAME"; > > >? > > Sure, that's the recommended way if can be sure that there is no chance > that that the count of rows will not change between the two selects. Keep > in mind that databases may be accessed by multiple users that may cause it > to be changed and the query results may noy be consistent if you don't take the > necessary precautions. > > On a side now, don't forget that count(rowname) may return a different > number than count(*) if row name has NULLs on it. > > >True, PHP doesn't handle the count, MS-SQL does. But it's guaranteed to > >be accurate (from whenever you executed the query plus a few > >milliseconds of lag time). > > If nobody else changes the database as noted above. > > >Then you don't need to worry about broken odbc_num_rows anymore... > > Notice that odbc_num_rows is not broken. It just returns what the ODBC > driver returns. Note that under Microsoft SQL, the COUNT aggregate function with a (*) argument returns the count of _ALL_ rows, not just NON-NULL values... The following should do as you ask, as long as you have M$ SQL. It will probably also work under variants of Sybase: BEGIN -- do a heavy duty lock on all tables involved in queries below, until the locks are released SET TRANSACTION ISOLATION LEVEL SERIALIZABLE SELECT COUNT(*) FROM TABLENAME WITH (HOLDLOCK) -- set locking back to default SET TRANSACTION ISOLATION LEVEL READ COMMITTED END