Fix collision calculation

[tskisner]

Small fix for intersections of circles with a multi-point segment. Closes #145.

[rncahn]

Excellent work, Ted. As I have said previously, it is important to check the entire collision code against something written independently. Bob

…

On Tue, Sep 25, 2018 at 9:21 AM Theodore Kisner ***@***.***> wrote: Small fix for intersections of circles with a multi-point segment. Closes #145 <#145>. ------------------------------ You can view, comment on, or merge this pull request online at: #146 Commit Summary - Fix bug in calculation of intersections of circles with a multi-point segment. File Changes - *M* src/collision.cpp <https://github.com/desihub/fiberassign/pull/146/files#diff-0> (14) Patch Links: - https://github.com/desihub/fiberassign/pull/146.patch - https://github.com/desihub/fiberassign/pull/146.diff — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#146>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AIeo3XflbIiCD161NWq5lZUzxYxCnk5Wks5uelgMgaJpZM4W4_SK> .

[sbailey]

Thanks! IIRC, Ted found this as a single discrepancy in 40k targets, which means that testing for collision consistency with other code (like whatever ICS will use) needs to use >>40k targets and/or very specific test dither patterns for confirmation.

[forero]

Thanks @tskisner . Sorry for replying too late. I was also checking against other runs to gauge the effect. This changes 1-5% of the target-fiber pairs in the final assignments on DR7.

[rncahn]

when I get a chance, I'll check to see whether the correction reduces the fraction of ELG's we get (or LRG, QSO).

…

On Tue, Sep 25, 2018 at 1:14 PM Jaime Forero-Romero < ***@***.***> wrote: Thanks @tskisner <https://github.com/tskisner> . Sorry for replying too late. I was also checking against other runs to gauge the effect. This changes 1-5% of the target-fiber pairs in the final assignments on DR7. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#146 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AIeo3TELLwBIRyxGh9G76Wan5OLOkZwhks5ueo62gaJpZM4W4_SK> .