I was just looking through my CF-Talk emails, when Tom Chiverton dropped a bomb shell on my brain. Someone had asked how to find out which function was calling a given function. Sounds complicated right (to me it does, so back off!)? But Tom was all like, oh, that's mad easy, just thrown an error and catch it. What?!? It's that easy?
I, of course, had to immediately test this, and yes, it is that easy. First, I set up a method called StackTrace:
<cffunction name="StackTrace" access="public" returntype="void" output="false"> <!--- Define local variables. ---> <cfset var Trace = StructNew() /> <cftry> <!--- Throw custom error. ---> <cfthrow message="This is thrown to gain access to the strack trace." type="StackTrace" /> <!--- Catch custom error so that page doesn't crap out. ---> <cfcatch> <!--- Create the trace object. The automatically generated CFCATCH object (from CFTry/CFCatch) should now have key elements that we can get our hands on. ---> <cfset Trace.StackTrace = CFCATCH.StackTrace /> <cfset Trace.TagContext = CFCATCH.TagContext /> <!--- Add stack trace to request. ---> <cfset ArrayAppend( REQUEST.StackTraces, Trace ) /> </cfcatch> </cftry> <!--- Return out. ---> <cfreturn /> </cffunction>
This code throws a custom error and catches it. The CFCATCH struct that is auto-generated by the CFTry/CFCatch block has the environmental context at the point the error was thrown. This includes the Stack Trace and the Tag Context. I wrap these up in a structure and append them to a REQUEST-scoped array.
Then, I created a few functions to test this out:
<cffunction name="A"> <!--- Get stack trace. ---> <cfset StackTrace() /> <!--- Return the return of B(). ---> <cfreturn B() /> </cffunction> <cffunction name="B"> <!--- Get stack trace. ---> <cfset StackTrace() /> <!--- Return the return of C(). ---> <cfreturn C() /> </cffunction> <cffunction name="C"> <!--- Get stack trace. ---> <cfset StackTrace() /> <cfreturn "C is the bomb!" /> </cffunction>
Then, I went ahead and tested it:
<!--- Create the stack trace array. ---> cfset REQUEST.StackTraces = ArrayNew( 1 ) /> <!--- Call all the methods. ---> #A()#<br /> #B()#<br /> #C()#<br /> <!--- Output the stack traces. ---> <h2> Stack Traces </h2> <cfloop index="intI" from="1" to="#ArrayLen( REQUEST.StackTraces )#" step="1"> <h4> Trace #intI# </h4> <cfdump var="#REQUEST.StackTraces[ intI ]#" /> </cfloop>
This gives us some great information. Take a look at the THIRD stack trace (the one produced by A()->B()->C()):
coldfusion.runtime.CustomException: This is thrown to gain access to the strack trace. at
It's not the easiest thing in the world to read, but you can certainly tell what functions call what functions. Even easier would be to look at the TagContext, but I am not in the mood to take screen shots... just trust me on that one. Anyway, Tom, you the man. Awesome suggestion!
hmm, not sure if I understood your post correctly, but wouldn't targetted cfdumps work just the same, and probably less code?
I do a lot of my deguggin with a cfdump followed by a cfabort. Then a cut and paste into the next part of my code till I find it.
Everybody does it differently I guess, whatever works for you!
You could also just dump the cf catch in the cf catch block.
<cf dump var="#cf catch#">
That's assuming the original question was directed towards debugging.
Let's hope the original question wasn't based on an idea of trying to build a function that works differently depending on the caller. Someone used to working with the custom tags and functions like GetBaseTagList() might think along those lines.
Good point. The only issue is that would only work in code that allows output to the page. Anything inside of a CFSilent tag or an output="false" function would not allow for this.
But as the guy above states, there are lots of ways to do this, I just had never thought of this one (the one I posted about). Whatever works for you, works :)
Ben, actually even if the code is inside cfsilent, etc., all you need to do is add a cfabort after the cfdump of the cfcatch object and you'll see it on the screen.
I am huge fan of the CFDump / CFAbort combo. The problem with it is that it haults the rest of the page. The beauty of CFTry/CFCatch solution is that you can execute the entire page and then get a full list of chained method calls / template executions.
Of course, this is only for use if you need the chained feature I suppose. If not, I whole heartedly stand behind the use of CFDump / CFAbort.
Couldn't you just use cftrace?
You could use CFTrace, however, it must log stuff and the user might not want to mess with the log files. Also, and I am not too familiar with CFTrace, but I think it outputs to the screen when it's called... where as this solution allows you to output to the screen at will.
Two valid solutions, two different solutions.
Ben, nice work (and kudos to Tom). You shouldn't have to jump to the defence of it though, any aditional method of debugging can only help and shouldn't be disgarded so quickly for preffered means. There's usually a time and place where debugging methods you notmally rely on don't help much, any additional tools can only be a good thing!
Good Job Ben...
hmmm...now I wonder who sparked that thread....oh wait...that's me ;-)
Well I guess my ramblings had to spawn a blog post somewhere sometime ;-)
Most definately. The more tools in the toolbelt, the better the building.
The more conversations we (as a community) get going, the more we are all gonna start learning :)
Absolutely Ben!! ;-)
Excellent post Ben. Just what I was looking for.
I have a piece of code in my application (it's a HUGE webapp) that is being called somewhere in the system and we just can not figure out what the hell is calling this function.
So I will most certainly be using this immediately!! :)
I think debugging is the perfect place for this.
You can find a stack trace tool that will throw a complete stack trace of your entire request to an external file at http://ketanjetty.com/coldfusion/monitor/
Does this help at all with the ability to determine at run-time, programmatically, the method that called the currently executing method?
You would not believe how many time Google refers me to your site for answers - thanks for all the effort you put into your blogs.
Based on this article, I have come up with the following to email me a stack trace
<cfmail to="#APPLICATION.sendEmailTo#" from="#APPLICATION.sendEmailTo#" subject="DEBUG #TimeFormat(Now(),'HH:mm:ss L')#" type="html">
<cfthrow type="Application" />
<cfloop array="#cfcatch.TagContext#" index="stTagContext" >
Looks good; may I ask why you need the stack trace? Typically, you would need that inside an error-handler; but in that case, you would already have an exception object to work with.
If it's a proprietary algorithm, however, no pressure.
I was trying to work out what was calling a certain function deep within a framework (Farcry CMS). I didn't want to bomb out as the function was getting called 5 times and I wanted to see the other 4 calls. The emailed stack trace shows me all the files called to get to the function.
Ah, gotcha. Glad it helped.
@Ketan Jetty -
Thanks for posting your monitor code. That is exactly what I was looking for - especially the example of how to use the GetDebuggingService. I'll be modifying it to select only those pieces I need.
This could probably be used to determine which templates are in use for a website as long as you had a way to run a full coverage robot.
Thanks for updating me on this. I am working on improving this monitoring service and hope to release a new version soon.
Not to post way after the fact, but I added an extra bit of code to make the stack trace easier to read, and appended it to a struct key so I could add it to whatever I'm working with. HTH someone out there! Thanks for all your posts, I have learned a lot from you, keep it up, man!
- <cffunction name="getStackTrace" access="public" returntype="struct" output="false">
- <cfset var functionTrace = StructNew() />
- <cfset var traceList = "">
- <cfthrow message="I want to see you, stack trace!" type="stackTrace" />
- <cfset functionTrace.stackTrace = cfcatch.stackTrace />
- <cfset functionTrace.tagContext = cfcatch.tagContext />
- <cfset traceList = replace(functionTrace.stackTrace, "$func", "!", "all")>
- <cfset functionTrace.TraceIDs = listToArray(traceList, "!")>
- <cfset arrayDeleteAt(functionTrace.traceIDs, "1")>
- <cfreturn functionTrace/>
Looks good. It's always nice to tie in more points of information in this kind of a situation. Certainly, I think have the tagContext is very nice as the stackTrace is... well, you see what the StackTrace is like to deal with.
Does anyone know how to get the name of the variable/instance of a CFC you're calling a method on?
I have x.cfc
I create 2 instances of it, one named i1, one named i2.
From within x.cfc.methodName, how do I get to the name of the instance executing it?
Not sure this is possible, if not I'm going to request it for the next release. Thx.
Just in case anyone else gets here searching for a similar term, ColdFusion added a new function in version 9 to do exactly this, get the current function name. GetFunctionCalledName()