I propose to change the syntax to
<inject_sources>
js=http://js1.js
js=http://js2.js
css=http://css1.js
<inject_sources>
If it’s in the main app the compiler will parse this and inject html, same way
it did before. Injected html is ensured to load serially, so it should load
before App.js.
If it’s in a module the compiler will parse this and inject a script to the
__deps file we create for modules, similar to what’s being done now. It may be
that we’ll need to add to the script some mechanism to ensure serial loading,
similar to what gcl does.
I can explore this when I have more time, probably next week, or someone else
is welcome to fix this differently.
Thanks.
From: Alex Harui<mailto:[email protected]>
Sent: Monday, May 18, 2020 7:53 PM
To: [email protected]<mailto:[email protected]>
Subject: Re: Script Loading Order (Continuing Heads-Up thread from Users)
Ah, ok. I thought that closure was injecting the main app's script block so
the dynamic script would be in front of it but it makes sense that wouldn't
happen at release time.
IIRC, we control the launch of the app. I thought there is one more script
somewhere that calls app.start() that is generated by the compiler. IMO, that
script can be altered to wait for scripts to load.
HTH,
-Alex
On 5/18/20, 8:44 AM, "Yishay Weiss" <[email protected]> wrote:
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%7C11eca28410ab4d377cb408d7fb4254da%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C637254134629069170&sdata=NuGReT8L9tH9K1VQNxgN%2BctVG3WDACYxMsaSywrTOSc%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.
