On 13 Dec 2014, at 7:46pm, James K. Lowden <jklow...@schemamania.org> wrote:
> Every DB Admin tool I've ever used proved to be more hinderance than > help. They seem to be written by the moderately competent to help the > novice, and run out of gas or fall over when faced with anything > complex. [snip] > > My first question, then, is whether or not the rowcount is so > interesting that it must be known before a table can be operated on. > I suggest the answer is No. The relative & approximate sizes of the > tables is known to the admin in most cases and, when it is not, the > information is readily discovered on a case-by-case basis. [snip] All true. Yet when I wrote my own DB Admin tool (suitable only for my own use, of no interest to anyone else) I included the same feature in it. When you click on a TABLE to select it the count(*) pops up along with information about the table's structure. I had no real idea why I put that in, it just seemed a natural thing to do. > That said, I'm puzzled why rowcount isn't maintained and exposed in > SQLite as part of a table's metadata, particularly when indexes/keys are > present. The cost of maintaining a rowcount is small, in terms of > computation and complexity. ISTM it is valuable information to the > system itself in evaluating query-plan costs. It does seem that knowing count(*) would be a very good thing to know for evaluating query-plan costs. I hope SQLite4 stores it. Simon. _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
