yandex.ru search engine is still being added

[shadowlmd]

Hello,

I'm using Chromium in Ubuntu, current version is 66.0.3359.181. Extension version 0.0.4 allows yandex.ru search engine to be added automatically after submitting making any search.

[cxw42]

Thanks for the report! We'll look into it.

[cxw42]

Yandex's search field already has a name - maybe that makes a difference?

<input class="input__control input__input" tabindex="2" autocomplete="off" autocorrect="off" autocapitalize="off" spellcheck="false" aria-autocomplete="list" aria-label="Запрос" id="text" maxlength="400" autofocus="" name="text">

[Procyon-b]

I have tried it too.
I think I know what is happening.
This page has 2 input fields. One normal with a name, the other hidden by default and without a name. It is used by the page to display "autocomplete" suggestion to the user. Since it has no name, like our textarea in its final form, it is not sent with the request.
I have made a modification in the code to ignore input fields with no name when it tests the form's structure (line 39 "if(elem.querySelectorAll...").
I don't know the syntax of -webkit-any() but it seems to work. Can you check if it's actually doing what I want: ignore unnamed form elements.

content.zip

[edit] by the way, here is a useful code to display the list of all forms and form elements of a page.
Paste this code in the javascript console of the page, and hit enter:

for (var i=0; i<document.forms.length; i++) for (var j=0; j<document.forms[i].length; j++) console.info({ i,j, html:document.forms[i][j].outerHTML, f:document.forms[i][j] } )

[Procyon-b]

It is definitively not correct.
It might break on other sites.
The syntax as I have used it is not doing exactly what I want

[cxw42]

OK- I will wait until I hear more before trying it myself :) . Welcome back! \o/

[Procyon-b]

This version should work.
Going back to CSS basics, and here is my take on it.
Modifications: some spellchecking in my comments. Comment out a line of debug output (line 24, too many info on aol.com, and duplicate of "unspoiled" array), removed output of unnecessary text and empty array (lines 69-70)

Here is the fix: (line 39)
old
':scope input:-webkit-any([type="text" i],[type="search" i],:not([type])):not([readonly])'
new
':scope input:-webkit-any([type="text" i],[type="search" i],:not([type])):not([readonly])[name]'

This means that we only take named input/text into account.

content.zip

Edit: I'll check if name="" in html tags is caught by the CSS selector. It shouldn't.

[Procyon-b]

I'll have to refine it to consider name="" as no name at all.
But it should work correctly except for some very specific cases.

[cxw42]

Looks OK to me on yandex, uk.webuy, and files.scene.org :D . I did some cleanup - would you please work off the latest on #17 if you do refine it? Commit 6164c20, and attached here. Thanks!

[Procyon-b]

Something is wrong with the cleanup.
I had to look for a couple of minutes before noticing that at line 47, the "===" breaks the code. Apparently when a variable is passed as is to "forEach()" (see line 68) it is transformed as an object when accessed through "this". If we want to use "===" we need to pass an object {now} instead of now and test on this.now ===

Edit;
PS: If you reinstate the debug line that I have commented out console.info({t:this, Found: elem}); (insert at line 24 of your code) you'll see how "this" ("t") is structured.

[cxw42]

Ah! Sorry about that! Thanks for enlightening me. I will amend that commit next time I am at my console. I like your idea of passing an object - since all objects are truthy (except null), the object==Boolean test may not have been ideal anyway.

[Procyon-b]

I have a fully functional version here.
Updated CSS selector, and using === with this.now

content.zip

[Procyon-b]

One last modification that we can do is increasing the timer delay. Push it up to 2 or 4 seconds, maybe.
This computer is a little bit slower and has less memory than the previous ones, so I ran into a "windows is swaping" event where chrome (I really had a large number of opened tabs) was taking a little bit longer than usual to reload uk.webuy.com. And this resulted in the timer firing too early and breaking the header of the page. It happened only once, and I haven't been able to reproduce it.

[cxw42]

Hmm... That's a tough one. Would you open a separate issue, please? We might need to make that one a user option, or request tabs permission and count tabs.

[Procyon-b]

No need to bother too much, 1.5 sec was chosen arbitrarily. We can select another value. It won't break anything at all.

[cxw42]

Should be fixed in #17 - closing. If you experience further problems, please feel free to comment back! Note that #17 hasn't been deployed yet, so you won't see the updates until @gregsadetsky rolls out v0.0.5.