royale-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alex Harui <>
Subject Re: [royale-asjs] branch develop updated: Fixes #258. But is that a proper fix?
Date Thu, 07 Jun 2018 19:13:06 GMT
I'm not sure I'm understanding.

There is no x,y in HTML/JS, so we can make it mean anything we want it to.  I hope the current
implementation tries to mimic Flash/Flex for backward compatibility reasons.  We could agree
to change that if we really want to, but I think backward-compatibility is useful here.

In Flex/Flash, if you set the x,y to 10,10, then the object is offset by 10 pixels from the
top-left of the parent.  If you read back x,y it will be 10,10.  However, in Royale, we map
x,y to the "left" and "top" styles.  if the parentNode has position=static, then we need code
to compensate for that.

One way is to make sure nobody has position=static.  That doesn't seem PAYG, might break snippets
from the internet, and can be overridden by someone setting position=static on an element
(not sure why anyone would do that).

Another way is, when you set x,y, we set position!=static on the parent.  I think we tried
that and there was some problem, but maybe we should try that again.  That would be PAYG,
IMO.  It is only applied when used.

But again, I want to understand the fundamental use cases.  The one you cited in RoyaleStore
turned out to be an un-needed hack.  What are the real use cases we need to consider?  How
important/prevalent is setting x,y outside of effects, popups, and absolute layout going to
be?   Otherwise, the code can be inefficient because you only pay for it in rare cases, which
is more PAYG then making every node pay for it "just-in-case".

My 2 cents,

On 6/7/18, 12:00 PM, "Harbs" <> wrote:

    I don’t think I was clear enough. The original issue that started this thread is actually
caused by the code which sets the y value based on the parentOffset. If the parentOffset is
ignored, the issue goes away and we don’t have to care about layout lifecycles.
    For the few cases where we need to read and set the *actual observed* x and y positions
based on the offsetParent which might be different than the actual parent, we can use utility
functions to get and set these values.
    > On Jun 7, 2018, at 9:27 PM, Harbs <> wrote:
    >> So, if we don't force position!=static throughout the DOM, then you have to have
code that compensates for that difference.
    > I don’t think I agree. Right now we’re modifying the x and y values because we
*might* care about the offsetParent. That’s not PAYG. In fact, the set values will be *wrong*
if the position of the parent is changed after the x and y values of the child are set.
    > Based on my observations, most apps will not need to set the values based on the
offsetParent, so hard-wiring that code in is not PAYG. This is especially true since setting
x and y currently forces a reflow of HTML. We’re suffering a major performance hit for no
    > In cases where we care about the parentOffset, we can use observedX and observedY
utility methods which account for offsetParent. That seems much more PAYG to me.
    > Removing the assumptions of reliance on offsetParent seems to eliminate all needs
to care about parent positioning.
    > My $0.02,
    > Harbs
    >> On Jun 7, 2018, at 8:23 PM, Alex Harui < <>>
    >> IIRC, the parentNode is always the parent of the child if you examine the DOM.
 offsetParent is the parent or grandparent, etc, that has position != static, and left/top/right/bottom
styles are always relative to offsetParent.  So, if we don't force position!=static throughout
the DOM, then you have to have code that compensates for that difference.
    >> IMO, the key issue is whether it is "ok" to force position!=static throughout
the DOM.  Can someone look at other JS frameworks?  I'll bet most of them use border-box like
we do.  If the major JS frameworks have opted for position!=static, then it might be the right
thing for us to do as well.  IMO, we would like to make it easy for snippets found on the
internet to work in Royale and they may not all presume position!-static.
    >> Also, IMO, our Containers should not presume position!=static.  Containers accept
assignable Layouts and the Layouts can set position!=static on the children and be appropriately
named (VerticalLayoutWithXYSupport).  That's PAYG to me.  Remember that TLCs should have very
little assumptions as illustrated in the ExplodedComponent example.  The beads can make assumptions
and be appropriately named and documented.
    >> My 2 cents,
    >> -Alex
    >> On 6/7/18, 6:15 AM, "Harbs" < <>>
    >>    I created a “simplify-position” feature branch which does away with the
offsetParent logic in UIBase. It does not change anything regarding position: static.
    >>    I have tested with my own app and a number of the examples. I haven’t found
any problems yet.
    >>    Input welcome…
    >>    Harbs
    >>> On Jun 7, 2018, at 12:20 PM, Harbs < <>>
    >>>> So, IMO, it would be nice to do a similar investigation of controlsPallette.
    >>> You are right. Removing the y value has no effect.
    >>> I am wondering that maybe it makes sense to apply relative to the Container
CSS selector and possibly a few others.
    >>> I’m trying to understand the specific cases where:
    >>>                if (positioner.parentNode != positioner.offsetParent)
    >>> Is required in setX, get x and setY, get y in UIBase. I would *really* like
to get rid of that code, and I’m, wondering what doing so would cause.
    >>>> On Jun 7, 2018, at 12:36 AM, Alex Harui <
<> < <>>>
    >>>> In the case of the controlsPallette, how did it get its size?  I could
certainly understand that if you didn't have position!=static, that setting top on the dockAndOuterContainer
would have no effect, but you shouldn't have had to set y or top in the first place.  IIRC,
you couldn't use x,y in Flex layouts like VerticalLayout/HorizontalLayout so migrating code
shouldn't be using it.  It is fine to create other layouts that support x,y as exceptions.
    >>>> In general, for a framework, we want to make sure we understand and fix
the fundamental problem before we address any hacks/exceptions.  IMO, the fundamental problem
in the scenarios you've provided so far is that the layout did not do what was expected so
someone tried using x,y to fix it.  First we need that layout do what is expected, then worry
about how folks might resolve other issues, if any.
    >>>> In ProductsView in RoyaleStore, the grip is an image loaded later, so
there might have been an issue there, especially on the SWF side, but I would expect the browser
to automatically re-layout once the grip image loaded.  I dug through Git history and found
that I was the one who hacked in the x,y.  It could be that early on, the layout did not use
FlexBox so we had a similar problem of responding to the grip image loading late.  But we
should remove the x,y and see if there is still a problem and ponder the right fix for that.
 ProductsView should not need to be setting x,y.
    >>>> So, IMO, it would be nice to do a similar investigation of controlsPallette.
 IMO, if you examine that div, it's offsetHeight should be 40 and if it is then you shouldn't
need to set on docAndOuterContainer which means that it shouldn't matter what
style.position is.
    >>>> My 2 cents,
    >>>> -Alex
    >>>> On 6/6/18, 2:12 PM, "Harbs" < <>
< <>>> wrote:
    >>>>> On Jun 6, 2018, at 11:05 PM, Harbs < <>
< <>>> wrote:
    >>>>> 		<js:Label x="20" y="20"
    >>>>> 				  text="{locStr.UPLOAD_YOUR_IMAGE}"/>
    >>>>   It actually, looks like the x and y values no longer have an effect
on this particular component, but there was clearly a reason they were needed to be specified
at some point…
    >>>>   Another one. I have an image which needs to stick to the bottom right
of the app. To do that I needed to following:
    >>>>     top: calc(100% - 21px);
    >>>>     left: calc(100% - 187px);
    >>>>     position: fixed;
    >>>>   With a default of position: relative, I’m able to do this:
    >>>>       top: -21px;
    >>>>       float: right;
    >>>>       right: 10px;
    >>>>   This being said, it actually looks like I’m wrong about the way to
set the defaults being .Application *{}. This actually has a *higher* specificity than .foo{}.[1]
    >>>>   I think the only way to guarantee that it’ll have a lower specificity
than other selectors is to use:
    >>>>   *{
    >>>>   position: relative;
    >>>>   }
    >>>>   I’m less happy about this option than ."Application *” because
it’ll effect elements outside the Royale app if it’s not in an iframe.
    >>>>   Harbs
    >>>>   [1]

View raw message