Remove infinite loop in linear polymer #3491
Conversation
Cap the number of retries for monomer placement to `max_tries`, as explained in the function documentation.
The `linear_polymer_positions()` function already has a retry mechanism with a default value of max_tries=1000. Running the test 10000 times with a random Mersenne Twister seed resulted in no failure. A single run required 3 retries in one test function, all the other runs placed a polymer on the first try in every tests. 6 runs took 1 min instead of 250 ms to execute.
Codecov Report
@@ Coverage Diff @@
## python #3491 +/- ##
=======================================
+ Coverage 87% 87% +<1%
=======================================
Files 536 536
Lines 24400 24254 -146
=======================================
- Hits 21250 21132 -118
+ Misses 3150 3122 -28
Continue to review full report at Codecov.
|
| @@ -186,6 +187,11 @@ draw_polymer_positions(PartCfg &partCfg, int const n_polymers, | |||
| } else if (not positions[p].empty()) { | |||
| /* Go back one position and try again */ | |||
| positions[p].pop_back(); | |||
| rejections++; | |||
There was a problem hiding this comment.
Can you please explain why this is needed? Couldn't you just increase attempts_poly?
There was a problem hiding this comment.
Ah sorry, when the first one has fixed position. But there has to be a more elegant solution than this...
There was a problem hiding this comment.
this while loop doesn't increment attempts_poly, therefore in high density systems, you can get stuck in an infinite cycle of deleting the previous bead and adding a new one at the same position
Description of changes:
draw_polymer_positions()(regression from Fix polymer generation #3484)polymer_linear.pytest that used to hide a bug in the previous implementation ofdraw_polymer_positions()