You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In the implementation of #55, the delimiter ', " or ` (ES6 only) for the packed string is chosen based on the shortest string after preprocessing (the one with the least escapes needed).
A tie between two is a common case, as most strings do not themselves contain another quote. Thus the same delimiter can be used for all of them, leaving the two others as equal options for the packed string.
The current tiebreaker is to avoid changing the delimiter of existing strings inside the code. Within the benchmark set, this solution gives gains in specific cases and no losses otherwise (compared to before #55).
Improvements are possible. The differences found among the benchmarks are :
Flappy Dragon : 977 with ` vs 978 with "
jscrush : 989 with " or ' vs 990 with `
Firewatch : 1029 with " or ` vs 1032 with '
This should be achievable by changing the tiebreaker of the cost function to account for the blocks that are formed / merged in the packer stage depending on the delimiter. A deeper analysis is needed.
The text was updated successfully, but these errors were encountered:
In the implementation of #55, the delimiter
'
,"
or`
(ES6 only) for the packed string is chosen based on the shortest string after preprocessing (the one with the least escapes needed).A tie between two is a common case, as most strings do not themselves contain another quote. Thus the same delimiter can be used for all of them, leaving the two others as equal options for the packed string.
The current tiebreaker is to avoid changing the delimiter of existing strings inside the code. Within the benchmark set, this solution gives gains in specific cases and no losses otherwise (compared to before #55).
Improvements are possible. The differences found among the benchmarks are :
`
vs 978 with"
"
or'
vs 990 with`
"
or`
vs 1032 with'
This should be achievable by changing the tiebreaker of the cost function to account for the blocks that are formed / merged in the packer stage depending on the delimiter. A deeper analysis is needed.
The text was updated successfully, but these errors were encountered: