OK, I looked at the commit for hljs, and the code it replaced. AFAICT, that is
an instantiation phase dependency and not a initialization phase dependency, so
it should not matter if it loads before or after app.js (unless someone does
use it in a non-lazy static initializer, which should be hard to do in Royale).
It should only matter that it is loaded before anybody calls it. Other than
static initializers, which should all be lazy, nobody should really call hljs
until after the application.start() is called in the index.html.
Here is the index.html for HelloWorld:
<html>
<head>
<meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1">
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<link rel="stylesheet" type="text/css" href="HelloWorld.min.css">
<script type="text/javascript" src="./HelloWorld.js"></script>
</head>
<body>
<script type="text/javascript">
new HelloWorld().start();
</script>
</body>
IMO, for applications that use inject_script (modules will use the _deps file),
we should generate code before the start() call that waits for any dynamic
scripts to load. So if HelloWorld needed hljs, the index.html would look more
like:
<head>
<meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1">
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<link rel="stylesheet" type="text/css" href="HelloWorld.min.css">
<script type="text/javascript"
src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js";
onload="highlight.min.js.loaded=true;"></script>
<script type="text/javascript" src="./HelloWorld.js"></script>
</head>
<body>
<script type="text/javascript">
var appInterval = setInterval(function() { if
(highlight.min.js.loaded) {
clearInterval(appInterval);
new HelloWorld().start();
}, 200);
</script>
</body>
Closure seems to use a hash of the URL instead of part of the URL to avoid
collisions in case two different scripts are called main.js or something like
that. And there might be some better way than using setInterval, but the idea
is to wait until the JS is loaded before calling start().
HTH,
-Alex
On 5/19/20, 12:18 PM, "Yishay Weiss" <[email protected]> wrote:
See 99a8c8356573ff16b668f2d39a447355c673fee3
Note that hljs is an externs file so I couldn’t implement static
initializers there.
There’s also a sort of a queue there for calls made before lib is loaded. I
realize this doesn’t scale as a pattern, which is why I proposed to simplify
annotations instead.
It could be of course there’s a simpler solution I’m missing.
From: Alex Harui<mailto:[email protected]>
Sent: Tuesday, May 19, 2020 10:03 PM
To: [email protected]<mailto:[email protected]>
Subject: Re: Script Loading Order (Continuing Heads-Up thread from Users)
Yishay,
I didn't think static initializers would require a façade or other fancy
mechanism. What kind of AS code ends up requiring this more complex solution?
-Alex
On 5/19/20, 10:34 AM, "Yishay Weiss" <[email protected]> wrote:
Hi Carlos,
Sorry for not responding earlier, I missed this post.
I haven’t been able to replicate this in debug mode, so it’s
interesting you’re seeing that.
I agree the façade solution is a bit cumbersome, but it works and maybe
it’s worth having it out there as an example of using static initializers
instead of injected code.
What do you think?
From: Carlos Rovira<mailto:[email protected]>
Sent: Monday, May 18, 2020 7:34 PM
To: Apache Royale Development<mailto:[email protected]>
Subject: Re: Script Loading Order (Continuing Heads-Up thread from
Users)
Hi Yishay,
I'm confused. The problem I reported was this;
ReferenceError: dialogPolyfill is not defined at
/Users/carlosrovira/Dev/Royale/Source/royale-asjs/examples/jewel/TourDeJewel/target/javascript/bin/js-debug/App.js:10:1
And just as I'm copying here I'm seeing that while I'm running
"js-release", notice that the link refers to "js-debug", so I think
there's
some wrong path involved here
I just updated with your latest change about hljs but I don't think we
have
a problems with it. A part from that I don't like the solution to make a
Facade for a script, since that involves to create 2 classes instead of
one. The solution should be just make 1 as3 file (instead of two) and
that
have the proper inject reference.
Please can you revert the hljsFacade?
thanks
El lun., 18 may. 2020 a las 17:44, Yishay Weiss
(<[email protected]>)
escribió:
> Unless I missed something that’s what it’s doing right now after my
fix.
> I’ll try to explain the scenario as I see it (no modules).
>
> Suppose we have an app that compiles to the following html.
>
> <html>
> <head>
> <script type="text/javascript">
> var script =
> document.createElement("script");
>
script.setAttribute("src",
> "hljs.min.js");
>
> document.head.appendChild(script);
> </script>
> <script type=”text/JavaScript”
> src=”App.js”></script>
> </head>
> <body></body>
> </html>
>
> After the first script element is loaded, the dom will look like:
>
> <html>
> <head>
> <script type="text/javascript">
> var script =
> document.createElement("script");
>
script.setAttribute("src",
> "hljs.min.js");
>
> document.head.appendChild(script);
> </script>
> <script type=”text/JavaScript”
> src=”hljs.min.js”></script>
> <script type=”text/JavaScript”
> src=”App.js”></script>
> </head>
> <body></body>
> </html>
>
> However, App.js will still be loaded before hljs.min.js because it
was not
> created dynamically. App.js will fail because it depends on hljs.
>
> From: Alex Harui<mailto:[email protected]>
> Sent: Monday, May 18, 2020 6:21 PM
> To: [email protected]<mailto:[email protected]>
> Subject: Re: Script Loading Order (Continuing Heads-Up thread from
Users)
>
> I don't think we have to inject these scripts into the main .js file.
The
> compiler knows when it is compiling the main app or a module. When
> compiling the main app, it should inject the script in the HEAD of
the html
> wrapper. For modules, it can inject the script into a separate file.
The
> ModuleLoader already loads extra files before loading the module. It
can
> load one more file.
>
> Of course, I could be wrong...
> -Alex
>
> On 5/18/20, 7:38 AM, "Yishay Weiss" <[email protected]> wrote:
>
> From what I’ve read [1] scripts injected dynamically will always
load
> after static script elements. So I don’t think there’s a good way to
ensure
> the proper order in run-time unless we do something like
> 99a8c8356573ff16b668f2d39a447355c673fee3 , but that’s verbose and
working
> with libs should be simple.
>
> Any ideas?
>
> [1]
>
https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.html5rocks.com%2Fen%2Ftutorials%2Fspeed%2Fscript-loading%2F%23disqus_thread&data=02%7C01%7Caharui%40adobe.com%7C755ef15c598f4953e07808d7fc296336%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C637255126994735409&sdata=ROGX%2BP%2Bm9Lln49Bml7HcXd4Ws0HM2nwM9UZfxCzR3oU%3D&reserved=0
>
> From: Alex Harui<mailto:[email protected]>
> Sent: Monday, May 18, 2020 8:03 AM
> To: [email protected]<mailto:[email protected]>
>
>
> Subject: Re: Script Loading Order (Continuing Heads-Up thread from
> Users)
>
> Every time I look, closure seems to change how it works. It looks
> like they are using callbacks and UIDs. I assume they can't use
await or
> Promise because of IE support. I haven't looked at the code you
generate,
> but might have to do something similar, IOW, wait for the callback or
known
> value before continuing.
>
> I think that if we create the script during the running of another
> script that we have to find a way to wait for that created script.
>
> It might help to know what kind of initialization code needed the
> definition so early. One alternative is that such code needs to be
> responsible for waiting.
>
> Most of our Application classes have a wait mechanism. We could
> leverage that, but that's also pretty late.
>
> It could be that for Applications we generate the script in the
head,
> and for modules we generate a separate script that is preloaded.
>
> HTH,
> -Alex
>
> On 5/17/20, 9:03 AM, "Yishay Weiss" <[email protected]>
wrote:
>
>
> >Is the script tag from inject_script going before or after
the
> script tag for the application (should be before, >IMO)?
>
> It’s going before but the network shows it’s loaded after.
>
> >Make sure the script tag has the same settings as the script
tags
> google closure uses in js-debug. I think they set some options so the
> scripts load in order.
>
> I see type being specified in the gcl script elements, while
> inject ones don’t. I suppose it’s worth seeing if that makes a
difference,
> though I couldn’t find evidence for that on the web.
>
>
>
>
>
--
Carlos Rovira
https://nam04.safelinks.protection.outlook.com/?url=http%3A%2F%2Fabout.me%2Fcarlosrovira&data=02%7C01%7Caharui%40adobe.com%7C755ef15c598f4953e07808d7fc296336%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C637255126994735409&sdata=K3j%2Fs5ztpCNBrEWl8M981ZgHZw4wCNWOXiRvYXbFJPE%3D&reserved=0
