Will Leshner wrote:
On Dec 20, 2005, at 7:46 AM, John Stanton wrote:
I haven't looked closely at the problem, so these are just first
ideas extending CM's approach. Basically there should be no reason
to perform any analysis of the SQL since that has already been done
and the metalanguage generated. My approach would be to prepare the
SQL statement and then use the compiled SQL to generate an UPDATE
statement. The existing SQLite VM gives you a model for the
framework of your engine.
This sounds like a very interesting idea. Could you be more specific
about how you would generate an UPDATE statement from the compiled SQL?
I have only browsed through the compiled SQL/VDBE so I cannot go into
detail, but the general idea is you should be able to have your own
interpreter which runs the compiled SQL and gives you what you want, a
list of table/column names for what is actually read from the database.
If any table/column is changed by the editor then you need to know the
rowid/table name/column name/new value and that is all you need to
construct an UPDATE, which would comprise one or more SQL UPDATE
statements. You would also identify calculated values and make them
non-changeable.
There may be some gotchas which would make the method difficult but you
would achieve the ability to read an arbitrary data set using any SQL
SELECT, edit it and write it back to the correct tables. The coding
should be relatively simple, but the analysis would be fairly arduous
but would leave you with a good knowledge of the SQLite VM.
JS
