> A delared type of NUMBER(10,2) has NUMERIC affinity, which means that > SQLite will attempt to store (string) values as integers first and floats > second before giving up and storing strings. >
This has nothing to do with my reply and I understand how it works. > > You do realize that there are decimal numbers that have infinite binary > expansions? > I wouldn't store such numbers into a NUMBER(10,2), just a NUMBER (I know they are the same in SQLite). > > You are also talking presentation (as in formatting) of numeric values as > opposed to representation (as in storing/retrieving). The former is best > handled in the user interface while the latter is the subject of database > engines. > My point was about not storing binary junk - the part of a number that has no meaning because the accuracy of the inputs is limited. When you have a generic db manager that can show any table or if you are looking at the results of your own SQL statement, it helps to reduce clutter on the screen. The data also compresses better. > > Fatihful reproduction of formatting would be possible using TEXT affinity > and calling sqlite3_bind_text. Performing arithmetic with these "numbers" > would however be tricky, slow and would still not guarantee that calculated > values would conform to the desired formatting. > > Of course, but in most cases we don't need to go this far. My main point is about rounding before binding; my secondary point that scale in a column definition can be desirable to avoid it. _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
