Signal prematurely cycling #12
Comments
|
I just pushed another commit to the branch; this is interesting: refresh==true gets all the way into type RefreshString = {ref :: Boolean, str:: String}
textInputWidget :: RefreshString -> Widget HTML RefreshString
textInputWidget rs = do
log $ "textInputWidget: " <> (show rs.ref) <> " " <> rs.str
txtNew <- D.input [valProp rs.str, P.unsafeTargetValue <$> P.onChange]
pure {ref: rs.ref, str: txtNew}
where
valProp = if rs.ref then P.value else P.defaultValue
textInput' :: Boolean -> CtrlSignal HTML String
textInput' refresh initVal = do
refstrNew <- sigNow refstr
pure refstrNew.str
where
refstr = {ref: refresh, str: initVal}
sigNow rs = step rs $ do
pure $ unsafePerformEffect $ log $ "refstr in textInput sigNow': " <> (show rs)
rsNew <- textInputWidget rs
pure $ sigNow rsNew
sig rs = do
pure $ unsafePerformEffect $ log $ "refstr in textInput sig': " <> (show rs)
debounce 500.0 rs textInputWidgetThis happens regardless of whether or not So it seems somewhat likely there is a problem with |
|
Hi @bbarker I will look at this (and the other issue you reported) over this weekend. |
|
It also occurs to me that I probably don't need |
|
I added a hack to workaround the issue. This is not a solution as you will see, but it does seem to work, so that may provide another data point. I manually modify the DOM after doing a serach for input elements. In the signal that creates a txtMay :: Maybe NonEmptyString <- D.div_ [P._id mjUI_dateInput]
$ textInput refresh (fromString prevTxt) -- assumes this id is unique ... (hack) :
pure $ unsafePerformEffect $ setChildInputByTag mjUI_dateInput "INPUT" prevTxtThe function is defined like: windowDoc :: Effect Document
windowDoc = do
win <- window
hDoc <- document win
pure $ HTML.toDocument hDoc
setChildInputByTag :: String -> String -> String -> Effect Unit
setChildInputByTag id tag value = do
doc <- windowDoc
let root = toNonElementParentNode doc
parEleMay <- getElementById id root
case parEleMay of
Just parEle -> do
let par = Ele.toParentNode parEle
childCollection <- children par
childArray <- HTML.toArray childCollection
let childMatches = filter (\c -> Ele.tagName c == tag) childArray
let childInputs = catMaybes $ HTML.fromElement <$> childMatches
-- log $ "Child array tags: " <> (show $ Ele.tagName <$> childArray) -- DEBUG
when (length childMatches == 0) $ log
$ "No children of " <> id <> " with tag == " <> tag
when (length childInputs == 0) $ log
$ "No input element children of " <> id <> " with tag == " <> tag
for_ childInputs (HTML.setValue value)
Nothing -> log $ "in setChildByTag, couldn't find element with id " <> id |
By prematurely, I at least mean before I expect it to. This may be working as intended, as I've not isolated exactly where the problem occurs. I have confirmed it is NOT an issue with debounce by disabling it. In the
dateInputbranch (commit = https://github.com/labordynamicsinstitute/metajelo-ui/tree/058ae0f942a0a5a43d164aebd75a108ff17664c2 ) I have a lot of log statements enabled.I confirmed that the value
refreshis set totruecorrectly when an XML file is loaded in the app (example file at https://labordynamicsinstitute.github.io/metajelo-web/), however,truenever fully propagates down to thetextInputsignal and associated widget, as can be seen from this log output:The end result is that, in some cases (namely dateInput, for some reason) won't update the value. I went down this route of adding
refreshbecause I noticed that fordateInput, usingP.valueinstead ofP.defaultValuewould allow it to be updated, but enabling that globally causes lag issues. Thus, I only want to do it whenrefresh == true(that is, when a user uploads a file) -- otherwise, we stick todefaultValue.The text was updated successfully, but these errors were encountered: