Regular expressions are the most useful tool in string manipulation going - but learning about them is a real pain. Essentially, they’re confusing for most folks because of an apparently opaque syntax, and because there are so many different ‘flavours’ of them.This book does more than anything else I’ve read to make all this clearer.

It’s a very ‘first principles’ book, looking at how regular expressions are processed, down to how they work through a string, character by character. This is useful, as it makes you realise how they work.

Later, it goes through more advanced features (lookaheads, etc.), and then it examines some of the most commonly used regular expression syntaxes - Java, .NET and Perl. It’s also got a lot of examples and good ideas on writing efficient expressions for the sort of stuff you do day to day - matching email, parsing arguments from the command line, that sort of thing.

At work, I’m our ‘regular expressions’ guy now, ‘cos I read this book, and I use it as a reference on a frequent basis. I also use regular expressions LOTS in my programming - and by God is it faster than writing code to process strings.

My only complaint - well, I wouldn’t mind a few wee examples in other languages - Javascript being the obvious one, and PHP and Ruby. They’re all VERY similar to Perl, so it wouldn’t take long to highlight any differences and explain how to use them.

However, it is a great book for the subject - just look at it on Amazon. My opnion? Every coder should read this book (or one just like it). Programming without Regexes is just crazy.

I could do this fine with the workflows that come with Sharepoint, but when I tried to run my own, it never worked. Whenever I added a new item, I’d get the message ‘Error Occurred’, and looking in the logs I found:

06/20/2006 12:00:28.75 OWSTIMER.EXE (0x064C) 0x0D08 Windows SharePoint Services
Workflow 72et Exception
System.ArgumentNullException: Value cannot be null.
at Microsoft.SharePoint.Workflow.SPWorkflow.GetReservedItemId( SPList list, Guid taskId, Boolean createNew)
at Microsoft.SharePoint.Workflow.SPWinOEItemEventReceiver.CreateListItemSubscription( Guid subscriptionId,
SPWorkflow workflow, SPList list, Guid channelTypeId, Guid staticCorrelationId, Guid dynamicCorrelationId,
Guid subEventType, SPEventReceiverType rt, String assemblyName, String className, String eventName,
Int32 itemId, Boolean fReserveId)
at Microsoft.SharePoint.Workflow.SPWinOETaskService.CreateSubscription(MessageEventSubscription sub)
at Microsoft.SharePoint.Workflow.SPWinOESubscriptionService.CreateSubscription(MessageEventSubscription sub)

I’m still not clear what was causing this - I couldn’t find more detail - but I did get rid of it for my troublesome workflows.

I found a thing suggesting that I’d have to have event handlers for all the events on the task that I create in this workflow - that is onTaskCreated and onTaskDeleted, as well as onTaskChanged.

I also found a comment suggesting that I’d have to have the line:

workflowId = workflowProperties.workflowId

in onWorkflowActivated.

Finally, I found a mistake in one of the settings for the correlation token on the onTaskChanged event.

I fixed all of these things, and the workflow worked on a new item being added. I removed the line of code from onWorkflowActivated. The workflow still worked. I removed the unnecessary event handlers - onTaskCreated and onTaskDeleted. The workflow still worked.

I can only conclude that the problem lay with the correlation token that I fixed. My working example can be downloaded here.

What I did then was go back and check the correlation tokens being used throughout the workflows I was having problems with. I’m still not sure how these work (though I do get what they’re for). For one of my workflows, I’d set this to the same variable name as the workflowId and taskId. This was wrong. I set them just to workflowToken and taskToken, and the workflow, well, worked. I checked out another workflow that was failing - and there, for the task, I was using a correllation token called workflowToken. I changed this to taskToken for things associated with the task. (I didn’t create a correlation token anywhere - it seemed to do that itself.) Now it seemed to work too.

I should point out, I don’t see where these correlation tokens are being used, and I really don’t get how they work. Yet.

Anyway, this seemed to fix the error I was getting above for all of my problematic workflows, although I got a new error for one - where I was trying to parse initiationData, but there isn’t any for a workflow started automatically like that.

This makes sense really - each cell is having it’s CSS rerun each time. Then it struck me - the styles were defined as:

td.frozen {
padding: 3px;
position:relative;
top: expression(document.getElementById('pane').scrollTop-2); /*IE5+ only*/
z-index: 5;
}

This mean that ‘getElementById’ was being run repeatedly. However, the style’s JavaScript was being run before ‘onload’. I just couldn’t run the ‘getElementById’ to populate a global variable after the element had been created, but before the style expressions were run. Instead, in a moment of clarity, I changed the style to:

td.frozen {
padding: 3px;
position:relative;
top: expression(getPane().scrollTop-2); /*IE5+ only*/
z-index: 5;
}

And added a script:

var pane;

function getPane() {
if( pane == null ) {
pane = document.getElementById(”pane”);
}
return pane;
}

Thus, we only run getElementByID once - the first time a CSS style’s javascript expression is run. This worked - the DIV tag now scrolls much more quickly, certainly not so as users will notice any lag.

Looking in the logs, I was getting an error ‘malformed header from script’. ‘Curious’, I thought. The full error was of the form:
malformed header from script. Bad header=isRed = 1:
/home/simplyxi/public_html/dispatch.cgi
It seemed to me that part of my code was appearing as a header when the page was returned.

Eventually, I figured it out. My code still had ‘puts()’ calls in it. I’d added these for debugging during development on WebBrick. In WebBrick, the strings output by ‘puts()’ were written to the command window WebBrick was running under. On my Apache installation, though, this was output as a header. Consequently, I got errors.

I removed the ‘puts’ calls, and it all worked fine.

Comments from my old blog:

Thanks, buddy. Saved my sorry ass.

By Magnus Bergmark at 18:07:07 Saturday 29th September 2007

Well, that solution works, although I guess it’s only useful in an intranet environment - it only works on IE. And it mystifies me why someone would want to or be allowed to run javascript from inside a CSS declaration. Still, we’ve checked it - works on IE 5.5 to 7.

I couldn’t figure out how to delete these characters - why is it that ‘vi’ is so popular? - so what I did in the end was create a new rails app and copy the cgi files across into my app. The files, although they appeared to have the same text, had very different file sizes.

Anyway, that seemed to do the trick.