Allow to trigger events on the view, update the app and make assertions on the view

[rogeriochaves]

This can only work after the PR #18 on elm-html-test is merged. EDIT: merged.

So, my goal with this is being able to test the views in a fully integrated way, since now we can get events from elm-html-test, we can use those events to trigger updates on elm-testable and assert things.

For the first iteration, I've just created a function where you have to query the html and call eventResult to return the Msg result from the node, this Msg is then passed to the update function, the API #1 looks like this:

test "triggers events" <|
 \() ->
     htmlProgram
         |> TestContext.updateWith
             (Query.find [ Selector.tag "button" ]
                 >> Events.eventResult Events.Click
             )
         |> TestContext.expectView
         |> Query.has [ Selector.tag "p" ]

This works fine, but it is kinda ugly, and I wanted to write something more natural, so instead of updateWith, I've created functions like query and queryAll that accepts a Query type from elm-html-test, the API #2 looks like this:

import Test.View exposing (..)

test "query for multiple nodes" <|
    \() ->
        htmlProgram
            |> query (find [ Selector.class "first-button" ])
            |> trigger Events.Click
            |> queryToAll (findAll [ Selector.tag "p" ])
            |> expectViewAll (count (Expect.equal 1))
, test "triggers multiple events" <|
    \() ->
        htmlProgram
            |> query (find [ Selector.class "first-button" ])
            |> trigger Events.Click
            |> query (find [ Selector.class "second-button" ])
            |> trigger Events.Click
            |> queryToAll (findAll [ Selector.tag "p" ])
            |> expectViewAll (count (Expect.equal 2))

So, this works, and at the first implementation I've made it in a way that it throws an error if something that was expecting a Multiple query sends a Single query and vice-versa. But on the next commits I've changed the types to tell that you are using a wrong function on compilation time rather than runtime.

But, this is still ugly, and queryToAll and expectViewAll are the worst names imo. So, I've decided to wrap the elm-html-test changed to the API #3:

import Test.View exposing (..)

test "triggers events" <|
    \() ->
        htmlProgram
            |> find [ Selector.class "first-button" ]
            |> trigger Events.Click
            |> has [ Selector.tag "p" ]
            |> count (Expect.equal 1)
, test "triggers multiple events" <|
    \() ->
        htmlProgram
            |> find [ Selector.class "first-button" ]
            |> trigger Events.Click
            |> find [ Selector.class "second-button" ]
            |> trigger Events.Click
            |> findAll [ Selector.tag "p" ]
            |> count (Expect.equal 2)

I loved it, it is totally readable. The downside if having to wrap all elm-html-test query functions (which are not very much).

The other downside/maybe unexpected behaviour, which also happens on API #2, is that after calling trigger the query is reseted, because when update is called it has to update the view, and the view lives inside elm-html-test queries to, so we have to reset it.

For example, on the last test of last example, after clicking the first-button the query is reseted and because of this I could query the second-button, which is not inside first-button as one might think.
At least personally this is not a problem, still think it reads very well, the "trigger" really does look like a break.

So, what you think?

[rogeriochaves]

The last commit can only be used if elm-html-in-elm PR #8 gets merged, if not, we can revert it

[avh4]

Great! (Will wait for eeue56/elm-html-in-elm#8 to get merged.)

I think the fully-wrapped Html.Query API is too heavy for elm-testable. I think I'd like to stick with

simulate : List Selector -> Event -> TestContext model msg -> TestContext model msg
expectView : TestContext model msg -> Html.Query.Single

I do think a cleaner API like you've proposed could be valuable, but I think that'd be better as a separate package that builds on elm-testable, and that elm-testable itself should strive for a small API rather than the most concise one.

[eeue56]

FYI I've merged all those PRs and published everything

[rogeriochaves]

@avh4 actually, for the simulate function I couldn't simply accept a List Selector because you may want to navigate deeper than that to trigger an event (find, children, etc).

So it end up like this:

simulate : (Test.Html.Query.Single msg -> Test.Html.Query.Single msg) -> Event -> TestContext model msg -> TestContext model msg
expectView : TestContext model msg -> Test.Html.Query.Single msg

The usage actually ended up very neat! I think I really liked more this approach:

htmlProgram
    |> TestContext.simulate (find [ tag "button" ]) Click
    |> TestContext.expectView
    |> Query.has [ tag "p" ]

But in case I change my mind, the other suggestions are in the revert commits, I might create another lib on top of elm-testable as you said later :)

[rogeriochaves]

Should we also change expectView signature so we can use it in a similar way to simulate? 🤔

htmlProgram
    |> TestContext.simulate (find [ tag "button" ]) Click
    |> TestContext.expectView (has [ tag "p" ])

[avh4]

Ah, cool!

I think leaving expectView as-is is good for now. I originally tried having it take a function like that, and it turned out to be ugly if you needed to do a more complicated expectation or if you wanted to do a few nested finds.

[avh4]

I think this looks good overall!

The only thing missing is some test coverage for the error cases in tests/ViewTests.elm. The three I can think of are:

• when the query doesn't find an element
• when the query finds multiple elements
• when the found element doesn't handle the specified event

[rogeriochaves]

@avh4 done! And thanks for that, indeed it was swallowing errors for not found events and elements.

If the query finds multiple elements the error is produced by elm-html-test the same way of when it finds zero elements, so I didn't include this and other extra cases cause I think we don't need to re-cover all the elm-html-test cases.

Also one nice thing is that the type system doesn't allow the user to simulate an event on a findAll for example.

[avh4]

"Failed to decode string" seems like a bad error message. Does a change to this message need to be made in elm-html-test?

[rogeriochaves]

Yes, this is due to this line: https://github.com/rogeriochaves/elm-html-test/blame/master/src/Test/Html/Events.elm#L123

I don't know why it is there, I'll make a PR to remove it, the error message is much better without it: "The event dblclick does not exist on the found node"

[avh4]

@rogeriochaves Thanks! I'm happy with this :D

I'm going to wait to merge until #9 (comment) is fixed in elm-html-test.

Also, FYI there are going to be some merge conflicts here, since I'm upgrading elm-format, and elm-test. But don't worry about those; I will fix them.

[rogeriochaves]

@avh4 done, it was not hard to fix the conflicts, telling git to ignore whitespaces during merge made things very easy.

[rogeriochaves]

doing this way to make tests less brittle to changes in error messages

[rogeriochaves]

@avh4 can we merge this?

[rogeriochaves]

@avh4 ping