Parsing shouldn't die when there are deselect problems about sc2reader HOT 5 CLOSED

The selection problems with the given hotkey group could also be temporary as well, further complicating this reporting mechanism. Group 1 might be quite accurate for 80-90% of the game and just have a short period involving these issues.

Every time they ctrl+hotkey to set the group we get 100% accuracy and existing issues with that group are resolved. ctrl+shift+hotkey on the other hand does not reset the group and persists the existing issues. It is pretty much a given that issues will slowly build over time but a player that frequently resets his control groups greatly mitigates this risk for us.

Perhaps each selection frame should have a error count that builds over time and resets when possible. In that way, really bad frames could (possibly) be thrown out.

from sc2reader.

I wasn't quite sure how to track and qualify selection errors but the above commit should remove any bad indexes before they can cause IndexErrors in the selection code.

We'll let the error reporting portion of this issue evolve as I better understand what is required.

from sc2reader.

Regarding error reporting, what would you suggest? I would rather review & approve your suggestion than try to come up with my own.

from sc2reader.

I'll do some iteration on ideas and get back to you.

Right now I'm running with a rough errors_detected/selections_made ratio as a rough guide. It works great for distinguishing problematic players but isn't nuanced enough to tell you what you can salvage (or what to throw out) from the player's selection.

from sc2reader.

sc2reader no longer dies when there are deselection mask issues.

Recent patches (1.5+) have also greatly improved our ability to track selections and removed most of original sources of error. As such, the need to distinguish between issues isn't so great and the player.selection_errors should probably suffice for now.

I'm suggesting this be closed, please do so if you agree.

from sc2reader.