Remove useless calls to ProductOpener::Display::remove_tags_and_quote in Import.pm

[stephanegigandet]

While investigating #2563, @zigouras found there were tons of calls to ProductOpener::Display::remove_tags_and_quote when we run import_csv_file.pl

Many of those calls are done on values that do not have defined values. The checks are now done only on defined values, and they are inlined in one place (thus removing the calls).

[stephanegigandet]

On a test import of 1000 products (with very few defined values):

Before: Profile of ./import_csv_file.pl for 52.4s (of 74.3s), executing 34385037 statements and 19654675 subroutine calls in 739 source files and 257 string evals.

After: Profile of ./import_csv_file.pl for 26.6s (of 35.6s), executing 12122014 statements and 2574614 subroutine calls in 739 source files and 255 string evals.

On the test import, there is a very significant speed improvement. The memory usage is almost the same though.

[zigouras]

While investigating #2563, @zigouras found there were tons of calls to ProductOpener::Display::remove_tags_and_quote when we run import_csv_file.pl

Many of those calls are done on values that do not have defined values. The checks are now done only on defined values, and they are inlined in one place (thus removing the calls).

Calling a function does not negatively impact performance by itself so putting the code inline will not make a difference. It is better practice to organize your code in functions.

[stephanegigandet]

Calling a function does not negatively impact performance by itself so putting the code inline will not make a difference. It is better practice to organize your code in functions.

Unfortunately in Perl calling functions does add overhead for each call, so we have to strike a balance between readability / maintainability in performance. The inlining is worth it only for functions that get called a lot.