onbeforeunload

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

Mark Peacock - 08 Sep 2003 17:42 GMT

To prevent accidental loss of lengthy text entry when a
user abruptly closes or navigates away from a page, we're
using

<body onbeforeunload="SaveTextEntry()">

to submit the form and execute server code before
unloading. In early tests we haven't been able to make it
fail; but is this the best way to accomplish this?

Thanks.
MP

Reply to this Message

Lewis Wang [MSFT] - 09 Sep 2003 03:56 GMT

Hi Mark,

I think it's not a bad idea to submit the form and execute server code by
"onbeforeunload". It works OK on my side, too.

If you need more information, just post it here, thanks.

Best regards,
Lewis
This posting is provided "AS IS" with no warranties, and confers no rights.

--------------------

| Content-Class: urn:content-classes:message
| From: "Mark Peacock" <markp@fb.com>
[quoted text clipped - 13 lines]
| Path: cpmsftngxa06.phx.gbl
| Xref: cpmsftngxa06.phx.gbl

microsoft.public.dotnet.framework.aspnet.caching:1397

| NNTP-Posting-Host: TK2MSFTNGXA11 10.40.1.163
| X-Tomcat-NG: microsoft.public.dotnet.framework.aspnet.caching
[quoted text clipped - 11 lines]
| Thanks.
| MP

Reply to this Message

Mark Peacock - 09 Sep 2003 14:38 GMT

Lewis,

thanks for your prompt response. The question has become
more complicated by more realistic tests (over the
production WAN) in which onbeforeunload fails
intermittently.

It seems as if client and server events only work
sequentially with each other by coincidence. So, if the
client event cycle of the page -- in this case, pre-
unloading, being re-submitted and finally unloading --
completes before the server-side code finishes, IE 5.5+
unloads and cuts short the code behind.

Is this an accurate understanding of what is happening?
How can we assure that server-side code that is fired by
onbeforeunload will complete? Or is there a better way to
accomplish this?

Thanks,
MP

>-----Original Message-----
>Hi Mark,
[quoted text clipped - 44 lines]
>
>.

Reply to this Message

Lewis Wang [MSFT] - 10 Sep 2003 09:24 GMT

Hi Mark,

I have done a simple test, but I can not reproduce this issue. If we submit
the web form in SaveTextEntry(), the server side will execute the code in
the Page_Load. Do you mean sometimes IE 5.5+ may cut short the code behind?
Would you please give me some steps to reproduce it?

You can write some logs at the end of Page_Load to confirm that server-side
code completes.

Hope this helps.

Best regards,
Lewis
This posting is provided "AS IS" with no warranties, and confers no rights.
--------------------

| Content-Class: urn:content-classes:message
| From: "Mark Peacock" <markp@fb.com>
| Sender: "Mark Peacock" <markp@fb.com>
| References: <3bda01c37628$3c7202a0$a301280a@phx.gbl>

<8vf#T4ndDHA.1996@cpmsftngxa06.phx.gbl>

| Subject: RE: onbeforeunload
| Date: Tue, 9 Sep 2003 06:38:02 -0700
[quoted text clipped - 10 lines]
| Path: cpmsftngxa06.phx.gbl
| Xref: cpmsftngxa06.phx.gbl

microsoft.public.dotnet.framework.aspnet.caching:1401

| NNTP-Posting-Host: TK2MSFTNGXA08 10.40.1.160
| X-Tomcat-NG: microsoft.public.dotnet.framework.aspnet.caching
[quoted text clipped - 75 lines]
| >
| >.

Reply to this Message

Mark Peacock - 10 Sep 2003 20:53 GMT

Lewis,

I appreciate your trying to simulate the problem. This
should help you see what we're seeing. (I sent you the
zipped solution by e-mail.) It uses Thread.Sleep in the
code behind to simulate server processing delay.

client script:

<HEAD><script language="vbscript">
Sub window_onbeforeunload()
Form1.btnSave.click
End Sub
</script></HEAD>
<body onbeforeunload="window_onbeforeunload()">
<form id="Form1" method="post" runat="server">

Code behind:

Private Sub Page_Load(..
With sb
.Append("User action:") 'Exit browser etc.
.Append("Event:") 'Page_Load
.Append("Server delay:") '1-5000 ms
.......
End With
LogFile(sb.ToString)
End Sub

Private Sub btnSave_ServerClick(...
LogFile("btnSave_ServerClick")
Thread.Sleep('1-5000 ms
LogFile("Finished server processing. ")
End Sub

The log output which apparently reveals the problem:
(According to other (unresolved) posts we've seen
elsewhere on this problem, a double postback is expected
with window_onbeforeunload; hence the duplicative log
entries.)

On the first run we see the complete, expected event cycle:
-The initial page load
-Subsequent page load fired by window_onbeforeunload
-Entering the btnSave_ServerClick event
-Finishing any server processing, after a 1 ms delay
(The server processing is on a fast intranet and likely
less than a second.)

-User action: Exit Browser; Event: Page_Load; Server
delay: 1 ms; 1:41:58 PM
-Event: btnSave_ServerClick; 1:41:58 PM
-Finished server processing. 1:41:58 PM
-User action: Exit Browser; Event: Page_Load; Server
delay: 1 ms; 1:41:58 PM
-Event: btnSave_ServerClick; 1:41:58 PM
-Finished server processing. 1:41:58 PM

On the 2nd run, even at 1 ms, none of the post-
onbeforeunload server processing completes.

-User action: Exit Browser; Event: Page_Load; Server
delay: 1 ms; 1:43:24 PM

A newsgroup tip we're using is to use <body
onbeforeunload="window_onbeforeunload()"> to call Sub
window_onbeforeunload() as this is thought to increase the
chance of getting the server code to run; which seems to
be confirmed (though not understood) in our tests.

The 3rd time we try it at 5000 ms, most of the
duplicative cycle completes, even with an artificially
long delay. It's enough to process our server code (which
we expect may run twice, in best case).

-User action: Exit Browser; Event: Page_Load; Server
delay: 5000 ms; 1:43:32 PM
-Event: btnSave_ServerClick; 1:43:32 PM
Finished server processing. 1:43:37 PM
User action: Exit Browser; Event: Page_Load; Server delay:
5000 ms; 1:43:37 PM
Event: btnSave_ServerClick; 1:43:37 PM

2 more times when no server-side processing occurs, even
with only a 1 ms delay.

User action: Exit Browser; Event: Page_Load; Server delay:
5000 ms; 1:44:29 PM

User action: Exit Browser; Event: Page_Load; Server delay:
1 ms; 1:45:13 PM

Finally, 2 in a row that behave as expected. Definitely
intermittent!

User action: Exit Browser; Event: Page_Load; Server delay:
1 ms; 1:46:35 PM
Event: btnSave_ServerClick; 1:46:35 PM
Finished server processing. 1:46:35 PM
User action: Exit Browser; Event: Page_Load; Server delay:
1 ms; 1:46:36 PM
Event: btnSave_ServerClick; 1:46:36 PM
Finished server processing. 1:46:36 PM

User action: Exit Browser; Event: Page_Load; Server delay:
1 ms; 2:01:20 PM
Event: btnSave_ServerClick; 2:01:20 PM
Finished server processing. 2:01:20 PM
User action: Exit Browser; Event: Page_Load; Server delay:
1 ms; 2:01:20 PM
Event: btnSave_ServerClick; 2:01:20 PM
Finished server processing. 2:01:20 PM

I hope this clarifies the problem. From what we've seen in
the groups, we won't be the only developers who would love
to get a solution to this.

Thanks again.
MP

>-----Original Message-----
>Hi Mark,
[quoted text clipped - 114 lines]
>
>.

Reply to this Message

Lewis Wang [MSFT] - 11 Sep 2003 10:45 GMT

Hi Mark,

After my research, I found that the reason for the intermittent POSTS is
due to a form that gets submitted asynchronously. In other words, closing
IE won't wait for the event to finish Processing.

I found the following workarounds:

1. Submitting the form in a hidden frame

You can have a hidden frame which does the actual submit. The main page can
then ask the frame to post it in its unload handler. If user closes the
browser, then there is no way to do this. You can open a new window, do the
submit and close it.

2. Doing a window.open and submitting the form there.

3. Prolong the onbeforeunload by doing a loop or something similar.

Hope this helps.

Best regards,
Lewis
This posting is provided "AS IS" with no warranties, and confers no rights.

--------------------

| Content-Class: urn:content-classes:message
| From: "Mark Peacock" <mark@fb.com>
| Sender: "Mark Peacock" <mark@fb.com>
| References: <3bda01c37628$3c7202a0$a301280a@phx.gbl>

<8vf#T4ndDHA.1996@cpmsftngxa06.phx.gbl>
<060001c376d7$97774c50$a001280a@phx.gbl>
<ytHprT3dDHA.2080@cpmsftngxa06.phx.gbl>

| Subject: RE: onbeforeunload
| Date: Wed, 10 Sep 2003 12:53:50 -0700
[quoted text clipped - 10 lines]
| Path: cpmsftngxa06.phx.gbl
| Xref: cpmsftngxa06.phx.gbl

microsoft.public.dotnet.framework.aspnet.caching:1408

| NNTP-Posting-Host: TK2MSFTNGXA13 10.40.1.165
| X-Tomcat-NG: microsoft.public.dotnet.framework.aspnet.caching
[quoted text clipped - 252 lines]
| >
| >.

Reply to this Message

Mark Peacock - 13 Sep 2003 00:02 GMT

Lewis,

I guess the idea of browser application, integrating
client and server processing, is still a stretch; but your
suggestion to prolong onbeforeunload with a loop seems to
do what we need.

Reading a hidden field, updated by code-behind, from
within the loop in onbeforeunload didn't work. Client
script never seemed to refresh the hidden form value to
catch the update. But it did notice an updated cookie
(updated by code-behind) and kick out of the loop to
finish unload.

Relying on cookies isn't perfect, but it's OK on IE-only
intranet.

In case anyone (like me, who is not a client script
expert) can profit by somewhat-tested code, here it is:

<script lang="javascript">
function GetCookie(sName)
{
var aCookie = document.cookie.split("; ");
for (var i=0; i < aCookie.length; i++)
{
var aCrumb = aCookie[i].split("=");
if (sName == aCrumb[0])
    return unescape(aCrumb[1]);
}
return null;
}
</script>
<script language="vbscript" id="clientEventHandlersVBS">

</script></HEAD>

Code-behind:

Private Sub Page_Load(..
Response.Cookies("Eval1").Value = "Start"
End Sub

Private Sub btnSave_ServerClick(...
' ... delay from server processing

Response.Cookies("Eval1").Value = "Done"

End Sub

Thanks for the help.
MP

>-----Original Message-----
>Hi Mark,
[quoted text clipped - 304 lines]
>
>.

Reply to this Message

onbeforeunload

Free Magazines