Program generation takes a significant amount of time

[redxeth]

In a recent project, we've found that as the flows get longer and more complicated that program generation due to finding or looking up unique test ids can be very time consuming. Even if you don't publish the Test IDS. This using local git archives.

It appears that there a lot of unnecessary reloading of the JSON in the latest release due to a previous bug fix to resolve range lookups. I have a branch that appears to resolve but still checking it out and adding a proper test case.

[priyavadan]

The reloading of the JSON file is not unnecessary IMO. The file has to be loaded to make sure that the pointers can be read and the next available number can be assigned.

Primary reason for my comment here was to ensure that a check exists for a database that is already published. Without it, (most likely?) the softbin ranges will get assigned from the first number in range instead of continuing to the next in range.

Edit: There is still a corner case bug though in the latest code, so I am going to update it in a bit.

[redxeth]

I dont have a specs case to cover an already published DB, would have to prob publish that here on github. But I do have a case using a dummy repo file, still testing it out.

If you can add a new issue with what you guys are seeing more recently as I think it's new (ruby version, etc.)