After some code review... The main problem with this code is the use of integers to maintain bit flags tracking cell usage. The advantage is that when the Sudoku search phase with its associated backtracking has to make a copy of the game board, the amount of data to copy is low. The disadvantage is that the code is unreadable. Trying to manage a bit mask with 81 bits using 64 bit integers makes it at least twice as bad.
But really the core of the algorithm is that it relies on 4 recursive CTEs. These are to construct the cell layout bit masks and digits, one to process the game rules, and one to do the backtracking search. They are crucial to the query. Currently Andl has recursive function calls but it does not have recursive CTEs. A recursive CTE is something special because rather than being true recursion it is actually a generator that populates a table with new rows according to a query. It does this by means of a queue rather than a stack. Andl could implement this algorithm using recursion, but it would be slow and might well produce a stack overflow, and would not be a good outcome. So the answer is: I plan to add 'recursive' queries to Andl. This may take a little while. Thanks for the challenge. If in the meantime you have any others that do not use recursive CTEs I would be interested. Regards David M Bennett FACS Andl - A New Database Language - andl.org -----Original Message----- From: sqlite-users-boun...@mailinglists.sqlite.org [mailto:sqlite-users-bounces at mailinglists.sqlite.org] On Behalf Of david at andl.org Sent: Tuesday, 9 June 2015 9:51 PM To: 'General Discussion of SQLite Database' Subject: Re: [sqlite] User-defined types -- in Andl Thank you. Exactly so. One of the problems with this kind of project is finding 'good enough' challenges to tackle. I'll let you know how I get on. [BTW I don't remember the last time I saw SQL like this. Understanding it might be the challenge...] Regards David M Bennett FACS Andl - A New Database Language - andl.org -----Original Message----- From: sqlite-users-boun...@mailinglists.sqlite.org [mailto:sqlite-users-bounces at mailinglists.sqlite.org] On Behalf Of Jean-Christophe Deschamps Sent: Tuesday, 9 June 2015 5:16 PM To: General Discussion of SQLite Database Subject: Re: [sqlite] User-defined types -- in Andl At 08:27 09/06/2015, you wrote: ???? >Andl is at a slightly higher level than SQL for writing simple queries. >Where it shines is writing complex queries that involve user-defined >types, custom transformations and custom aggregations. For complex >relational operations there is nothing I know that can come close, >productivity wise. `--- You call for complexity, you get it! What I would find pretty convincing is seeing how andl would translate the self-contained sudoku solver posted by E. Pasma in this post: http://mailinglists.sqlite.org/cgi-bin/mailman/private/sqlite-users/2014-Mar ch/051982.html Granted, this is far from the typical SQL you can find in routine use, but I believe that andl being able to elegantly translate it would certainly impress a number of readers and make many of us more interested in digging further. I'm not throwing such a mayhem challenge at you in the tone of "if andl can't do that, then shut up". Maybe andl is not yet complete enough today to achieve that and this wouldn't be a big issue. But if it can I'm sure andl will attract more attention. JcD _______________________________________________ sqlite-users mailing list sqlite-users at mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users _______________________________________________ sqlite-users mailing list sqlite-users at mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
