@raster if I understand your comment well, I think the (sd->text) does not always have the entry text, specially if you append text in-front or back, so condition will be on "elm.text" part to ensure this is entry latest text

developer should not do case 1 it is invalid, but it case he did it, we will help him :)

use goto label

move logic to edje_entry

@raster again I am not against caching, caching is great.

Normally inside destructor you will check all data members for your object. (so you will not forget them), but for global maybe you did not know that you have global ones.

@bu5hm4n It is bad Idea because it lead to :
1- Hard to follow memory leaks,
2- With any updating to the code,There is high probability to accidentally forget about global vars ,(This case shows my point) (like this memory leak)

It work if we are going to set text in "elm.text", else we do not care inside this function

add event thaw

move code, and align with code convention

code convention

follow EFL code convention for while/if statments

• fix typo in the comments

@segfaultxavi I already update them Test Plan

Thanks for all the information

-search common escapes first
-EFL coding conventions

@segfaultxavi I will update the code to get better results that old one for pure common escapes.

I think it is make sense to reuse memory, instead of realloc

• fix typo in the comments

Old Code
total time for escape_string is 2.370570
total time for common escape_string is 0.100057
total time for string_escape is 2.430365

@segfaultxavi Yes I think we can skip first '&' and last ';' from checking, this will enhance performance more

@raster Thank you for clear it out.

@raster Can you please take look at this?

99% of the time, it is a bad idea to use global variables inside library

remove merged changes

update

@vtorri I will check if it will effect performance.

In D8497#154188, @bu5hm4n wrote:

Hi, i can see that this giantly impoves the time spent in this function. So this is fine from my side. However, lease note that if you want to fix performance problems in EFL that such tiny functions will not matter that much in total. As @zmike pointed out, there is a other function which is a bit more wastefull in this regard.

Further more, running entry examples with perf for roughly a second does not even bring this function up for me. So yes this is a performance improvement of this function, but no, it will not matter much for the overall performance of EFL. I don't know if the overall performance of EFL was your goal.

In D8497#153974, @zmike wrote:

Your test case should probably use a string that has some real <br> and <ps/> tags?

update code to use strchr

In D8497#153949, @segfaultxavi wrote:

Also, any performance-improving patch should be accompanied of numbers (and how to obtain them). Otherwise, how do you know if you're actually speeding up anything?

In D8497#153941, @zmike wrote:

Seems like this could be simplified a bit? And if you're looking to optimize this kind of thing then probably you'd also want to improve performance of _entry_remove_new_line since this function does a full copy of the string and then 4 full iterations over it?

this function return count of selection regions, not the length of selected string

undo some old commits

Before Fix

I set font size just to show the effect without writing a lot of text (instead of many glyphes we have huge one ),
and effect of copy will increase when image size increase.

On my Machine i7-2600 Ubuntu 16.04 LTS

This will have small increase effect in performance (still may be it is good to have )