On Feb 29, 2012, at 10:28 PM, Richard Hipp wrote:
> But that would require a lot of
> extra code space, which is not "lite".
There is a fine line between being parsimonious and cutting corners.
For example, one can define a named unique key constraint:
constraint header_uk unique( name ),
This will result in SQLite automagically creating an unique index to support
the constraint. What's the name of that index? No one knows for sure, as it's
automatically generated (e.g. sqlite_autoindex_header_1 or something) instead
of simply using the provided name of the constraint.
This might not have mattered much in the past as one couldn't refer to indexes
by name.
But now SQLite provides the INDEXED BY phrase [1]. Where one can refer to
indexes by name. Index names one has no ideas about, as SQLite took it upon
itself to randomly generate a totally unrelated name.
Sigh…
It's my opinion that SQLite does itself a disservice by systematically ignoring
all the meta information painstakingly provided in the DDL in the first place.
Just my 2¢.
[1] http://www.sqlite.org/lang_indexedby.html
_______________________________________________
sqlite-users mailing list
[email protected]
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
