Incremental Blob Support

[mgd134]

Hi,

I'd like incremental blob support in node-sqlite3.

https://www.sqlite.org/c3ref/blob_open.html

I have forked the repo and added a prototype of the functionality that currently works. I'm not quite sure what to make of the concurrency framework (uv_work_queue stuff) and the callbacks, however. I'm happy to finish it if someone can point me in the right direction and if you guys wouldn't mind merging it when it is done to your satisfaction.

I didn't want to create a pull request quite yet due to the issues noted above.

Thanks for any guidance,
Mike

[springmeyer]

If you have specific questions about the uv_work - please ask.

If you are keen to submit a pull test coverage will be important.

[mgd134]

OK, I just want to be sure I understand correctly how the concurrency stuff works. Here's the way I'm reading the code, using Database::Exec as an example:

The NAN_METHOD(Database::Exec) is used to make sure the user passes in the correct parameters, create the baton, and schedule the Work_BeginExec function to run at some point later.

The Work_BeginExec method is used to do a few sanity checks and then put the Work_Exec method on the queue. This schedules Work_Exec to be run on a new thread and then Work_AfterExec to run when Work_Exec finishes, correct?

The Work_Exec method actually runs the sqlite3 command.

The Work_AfterExec method checks for errors and calls the callback if one exists.

Do I have this understanding correct? How do I decide if I need all of these separate steps for the APIs I'm adding?

I can add some test cases for the new API.

[springmeyer]

The Work_BeginExec method is used to do a few sanity checks and then put the Work_Exec method on the queue. This schedules Work_Exec to be run on a new thread and then Work_AfterExec to run when Work_Exec finishes, correct?

correct.

Do I have this understanding correct? How do I decide if I need all of these separate steps for the APIs I'm adding?

You need those separate API's if you want to avoid blocking the main loop.

If you have a pull request, queue it up for others to play with. Be clear about what it makes possible!

[daniellockyer]

This would be nice but the issue has gone stale so I'm going to close for now - please reopen with an issue template if it's still something that is desired