add keyword Set Selector Suffix

[tkopra]

Is your feature request related to a problem? Please describe.
Currently I need to separately add ">> visible=true" at the end of each selector, especially text selectors. This seems unnecessary repetitive work and also makes the test less clean.

Describe the solution you'd like
New keyword Set Selector Suffix. Similar to existing Set Selector Prefix.

Describe alternatives you've considered
More generic Set Config keyword to handle all similar configuration changes.

[tkopra]

@aaltat is there a chance that this would be implemented in next few months?

[aaltat]

There is always a chance if you want me to implement it, but odds is not that high. Would you like to implement it yourself?

[tkopra]

Ok thanks for the update. I am not really python coder so cannot help with the implementation, sorry. I can help with testing and documentation though, if that helps?

[aaltat]

Well, writing some acceptance tests for the feature that demonstrates how the feature should work is always a good idea. It can be in a PR or you can add them to the issue too

[tkopra]

I was thinking about manual testing. My experience with writing unit tests is close to zero. But I can perhaps try if you give some guiding?

[aaltat]

I am not talking about unit tests, but Robot Feamework acceptance tests, like in here https://github.com/MarketSquare/robotframework-browser/blob/main/atest/test/02_Content_Keywords/secret.robot

[tkopra]

Ok. I cantry but need more assistance for that. I am not sure how to test this at Browser keyword level. I tried to find acceptance tests for Set Selector Prefix, but did not find?

Also there might be a logical issue with Set Selector Suffix with the idea to "automate" visibility check. nth is also needed very often and that is dynamic, not static. Problem is that there is often need to do "..>>visible=true>>nth=${nth}". With Set Selector Suffix approach this would become ">>nth=0>>visible=true" which is not necessarily the same element anymore I believe?

[aaltat]

There are text example here: https://github.com/MarketSquare/robotframework-browser/blob/main/atest/test/01_Browser_Management/element_selector.robot#L162-L172

Also the scope feature should be tested, which is in https://github.com/MarketSquare/robotframework-browser/tree/main/atest/test/08_Scope_Tests

[tkopra]

Did I understood correctly that there are static html pages inside the repo which can be used to test the keywords? I could add there new page with a hidden element (if not already existing one) and test the visibility suffix with that?

But before going into that, did you understand my concern for the order of the visible=true and nth=N? And any ideas if that can be solved? I mean that if I define visibility=true suffix, and use nth in selector, nth would take into account also hidden elements because suffix is applied last? And would not necessarily give the desired outcome (which is taking nth visible element).

[aaltat]

1. Yes we have lots of test pages https://github.com/MarketSquare/robotframework-browser/tree/main/node/dynamic-test-app/static
2. Yes I did see the question, but I don’t know the answer. Most likely easiest way is to create a PoC and try it out, unless this is clearly stated somewhere in the Playwrights documentation.

[tkopra]

I tested my concern for the order of the visible=true and nth=N. Unfortunately there is an issue.

For example, these two are not equivalent:
//input >> visible=true >> nth=2
vs
//input >> nth=2 >> visible=true

In the first case, indexing is done among visible elements.
With a global suffix, indexing happens before visibility filtering, in DOM order, so hidden elements can still affect which element is selected.

Knowing this limitation, Set Selector Suffix would unfortunately not be a usable solution for my main need, and addressing that need would likely require a different approach altogether.

Any ideas?

[aaltat]

I did peak our code and we add the nth=n in few cases, in practice when looping trough elements and when string mode is disabled (Playwright has strict mode always on and it can not be configured off.) All elements are always find trough findLocator, the code is somewhat complicated; https://github.com/MarketSquare/robotframework-browser/blob/main/node/playwright-wrapper/playwright-invoke.ts#L31 It can made simpler, but the problem is difficult to solve so that we could tweak the locator finding to //input >> visible=true >> nth=2.

Your problem is difficult to solve. it is easier to solve if strict mode or looping is not used. But then it is not complete solution that would work in all cases. Also it comes really hard to explain and then it will come difficult to understand.

I am open to good ideas too.

[tkopra]

Ok thanks for the update. I try to install Playwright too and see if I get any ideas from testing that in similar cases.