On Wed, Sep 6, 2017 at 7:56 AM, Hick Gunter <h...@scigames.at> wrote:
> Count(<name>) needs to extract the <name> field from the record, tallying > only those that are NOT NULL. > Technically it would not need to "extract" the <name> field, only lookup the row header and see whether that field/column is NULL or not (since NULL is a special SERIAL TYPE of value 0 [1]). asql> explain select count(a) from t; > addr opcode p1 p2 p3 p4 p5 comment > ---- ------------- ---- ---- ---- ------------- -- ------------- > ... > 4 OpenRead 0 2 1 1 00 t > 5 Rewind 0 9 0 00 NULL > 6 Column 0 0 3 00 t.a > 7 AggStep 0 3 1 count(1) 01 NULL > ... > But as you showed, in practice it may still do, because of the Column VDBE op-code. Unless one imagines a specialized ColumnNullness of ColumnType new op-code that wouldn't need to extract the actual value/content, as a future optimization. --DD [1] https://www.sqlite.org/fileformat.html#record_format _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
