Support for frames/iframes

[aik099]

Idea is to allow specifying frame/iframe name where element is located so that when somebody wants to access it an automatic switch to that frame happens and then switch back to original window/frame happens.

Possible implementations:

• have inheritable @frame('test-name') or @frame('name' => 'test-name') annotation
• have frame parameter for @find-by annotation (e.g. @find-by('id' => 'test-id', 'frame' => 'test-name'))

When used then:

• current frame will be changed using switchToFrame method of Mink
• a search will be made for an element
• current frame is changed back to what it was

If was used on element container (sub-class of AbstractElementContainer, that was called HtmlElement before), then there is no need to specify frame/iframe on each element in this container because they'll be searched within container located in another frame/iframe anyway.

Egde cages:

• nested framesets - I think Mink should be handling that already
• frames/iframes in another window/tab - I think it's out of library scope because user can switch window before doing element search

[evangelion1204]

@frame sounds better to me else @find-by might get too complex and an extra annotation is easier to spot.

[evangelion1204]

Just thinking about a possible implementation should we either enhance the DefaultElementLocator or create a new FrameElementLocator? The Factory could either return a Default or Frame depending on Property

[aik099]

The default element locator can handle that. Nobody really restricts one element locator to operate only on one annotation.

[evangelion1204]

In that case we need to access either the session or the driver itself to switch to Frame. Either we put it as part of the Locator or as part of the ISearchContext.

[aik099]

No need. The search context is a real element (if I'm not mistaken), not a proxy. When we switch a frame and find an element there, then the returned element reference (inside Mink) contains (that needs to be tested) the frame where element was found. So even if you switch the frame, the element retrieved earlier should still act as being inside a frame and all subsequent findAll calls from locator should honor that.

[evangelion1204]

I tested it, the element is found. But accessing it and changing the value will force selenium to search for the element again to alter the value resulting in a not found when switching back initially.

This means that we either need to hook on every driver access (location of the element) and switch the target frame or the implementation itself needs to be done in Mink and its drivers.

[aik099]

You're correct. Mink only retains XPATH of element found via the Driver (e.g. Selenium, Sahi). Then this xpath is used to perform all further actions (even finding sub-elements).

To make this happen we need to:

1. store frame handle (driver specific frame identifier) next to WebElement
2. before every operation switch frame to needed one (if they differ)
3. after every operation switch frame back (if needed)

Doesn't sound like an easy task. Let's postpone this for now.

[aik099]

Interesting idea posted here: yandex-qatools/htmlelements#56 (comment)

So we switch to frame once > do the stuff > switch back. Since we do it outside of the elements no need to switch back/forth on each element query.