I have a SERVER variable, a structure of structures, which is highly accessed by all users (concurrency). Values are added and deleted very very frequently. Here is a small example of my SERVER variable. The real one has much more data.
<cfset SERVER.structure = StructNew()>
<cfset s = StructNew()>
<cfset StructInsert(s, 'test11', 'value11', true)>
<cfset StructInsert(s, 'test12', 'value12', true)>
<cfset StructInsert(SERVER.structure, 'test1', s, true)>
<cfset s = StructNew()>
<cfset StructInsert(s, 'test21', 'value21', true)>
<cfset StructInsert(s, 'test22', 'value22', true)>
<cfset StructInsert(SERVER.structure, 'test2', s, true)>
Every couple of hours, I loop this structure to clean expired data. However, I am getting the error "null null" while looping the variable like this:
<cfloop collection="#SERVER.structure#" item="key">
<cfif StructKeyExists(SERVER.structure, key)>
<cfloop collection="#StructFind(SERVER.structure, key)#" item="key2">
<!--- And some code here --->
</cfloop>
</cfif>
<cfif StructCount(StructFind(SERVER.structure, key)) eq 0>
<cfset StructDelete(SERVER.structure, key, false)>
</cfif>
</cfloop>
I'm receiving the error in the first line of the example. In this line, exactly:
<cfloop collection="#SERVER.structure#" item="key">
So I tried another approach. Instead of looping one by one, I created an array of keys and looped it. Unfortunately, the "null null" error was also happening there, in this exact line:
<cfset arrayOfKeys = StructKeyArray(SERVER.structure)>
My first theory is that ColdFusion can't handle the concurrency level that this SERVER variable has. I tried to use a <cflock> here, while clearing the variable, but it didn't work either. And I can't have this <cflock> where the variable is actually being used and modified by users because of the extra load that it would add (I believe).
I don't know... I'm out of ideas. Does anyone now why is this happening or have suffered this problem before? And a solution or workaround to this problem, or even a suggestion to make my code better, would be more than welcome too.
Thank you very much.
I tried to use a <cflock> here, while clearing the variable, but it
didn't work either. And I can't have this <cflock> where the variable
is actually being used and modified by users because of the extra load
that it would add (I believe).
This is your problem. If you are using the server scope, you must lock all access to it (read and write). Otherwise you will get errors. That's the long and the short of it.
My first theory is that ColdFusion can't handle the concurrency level that this SERVER variable has
Well: no. ColdFusion will synchronise individual operations to the server scope (well: it's handled at Java level), and that's where its job begins and ends. It's just that your approach doesn't handle it. Namely that you don't take steps to mitigate race conditions in your own code. This assertion of yours:
I am not locking access when I work with the variable because collisions can't happen
Is just wrong. There's a race condition in your loop.
As others have hinted at, this is very poor application architecture, and the code is facile.
Just put the data in a database. That's what databases are for, and they're written in such a way to optimise the sorts of operation you are trying to (but not succeeding, obviously) to achieve here.
I think this is probably a case of premature optimisation: have you had this data in a properly provisioned and designed DB? Or are you second-guessing that it will be a problem? I suspect it's the latter. Or that the DB server was not properly provisioned.
Over and above the database tier, you could also use a caching tier like others have suggested.
But just don't reinvent the wheel, like others have said.
The bottom line answer to your question is that you're getting the errors because you are not correctly locking the data, and you have race conditions in your code, making ColdFusion trying to access data that you have told it is there, but could subsequently have been changed. This is due to your code, not a shortcoming in ColdFusion.
Related
I'm calling a web service through ColdFusion which returns an object, and I want to verify if one of the methods of this object exists as it won't always exist.
I found this source which seemed promising however based on my tests I can see the results are always negative and the method is never found when it's clearly there.
<cfif structKeyExists("#Result.getNotifications().getValidationResult(0)#","getField")>
Result is my underlying object, and my end goal is to verify if the method getField() exists.
Is there a clean way to do this as opposed to a try/catch?
Update:
Unfortunately, I am not sure IsInstanceOf() works with web services, due to the fact that CF uses a Proxy object to "wrap" the underlying web service class. If not, another simple option is to check class name. That avoids the ambiguity of checking for method name only (which could potentially exist in many different classes). Plus I suspect it may be more light-weight than IsInstanceOf() anyway.
<cfif compare(yourObject.getClass().name, "org.tempuri.ValidationResultField") eq 0>
Found ValidationResultField. do something
</cfif>
It looks like the dump contains several different types of objects/classes: ArrayOfValidationResult, ValidationResultField, etecetera. It sounds like what you are really trying to determine is which of those classes you are working with, so you know exactly what fields and methods will be available, per the web service definitions. Given that, I think IsInstanceOf() would be a more appropriate test, than checking for method names. More accurate as well. Nothing prevents two different classes from having the same method name. So even if method X or Y exists, there is still a possibility it may be a different class than expected.
<cfif IsInstanceOf(yourObject, "org.tempuri.ValidationResultField")>
do something
</cfif>
As far as I know, the mentioned structKeyExists approach only works if CF wraps the class internally, e.g. all instances of cfcomponent.
The only option left is to actually reflect the class:
<cftry>
<cfset Result.getNotifications().getValidationResult(0).getClass().getMethod("getField", javaCast("null", ""))>
<!--- method does exist --->
<cfcatch type="coldfusion.runtime.CfJspPage$UnsupportedBaseTypeException">
<!--- method does not exist --->
</cfcatch>
</cftry>
If the method doesn't exist, it throws UnsupportedBaseTypeException, which seems to be a follow-up of NoSuchMethodException.
Honestly, you might as well just invoke the method and catch it. Reflection comes with an additional overhead and you have to catch it anyway.
Like Miguel-F, I think this is something for getMetadata(). The following should return an array containing the respective functions of the object:
<cfset funcs = getmetadata(nameOfObj).functions>
The names of the functions are then funcs[1].name, funcs[2].name, and so on.
In general, you may obtain the metadata of all the functions of a webservice, given the URL of the WSDL, with something like
<cfhttp method="get" url="http://www.webservicex.net/globalweather.asmx?WSDL" result="res">
<cfset wsXml=xmlparse(res.filecontent)>
<cfset wsOperations = xmlsearch(wsXml,"//wsdl:operation")>
<cfdump var="#wsOperations#">
Another method you could look at (perhaps undocumented) is to get the method names from the class names in the stubs directory.
The code to run is:
<cfscript>
wsargs = structnew();
wsargs.savejava="yes";
</cfscript>
<cfset convert=createobject("webservice","url_of_wsdl",wsargs)>
Then figure out how to fish out the names from the stubs directory, {CF_INSTALL}/stubs. In my case, CF_INSTALL is C:/ColdFusion2016/cfusion/
Is there some way to improve these server-side user fields validations?
<cfif Form.LoginName EQ ""><h1>Login Name is required.</h1></cfif>
<cfif Form.Password EQ ""><h1>Password is required.</h1></cfif>
<cfif Form.Password NEQ Form.PasswordConfirmation><h1>Password confirmation does not match Password.</h1></cfif>
<cfif Form.FirstName EQ ""><h1>First Name is required.</h1></cfif>
<cfif Form.LastName EQ ""><h1>Last Name is required.</h1></cfif>
<cfif Form.LoginName EQ "" OR Form.Password EQ "" OR Form.Password NEQ Form.PasswordConfirmation OR Form.FirstName EQ "" OR Form.LastName EQ "">
<p>User has not been created</p>
<p>You can use your browser's back button to keep form fields filled and try again.</p>
<p>Return to users list.</p>
<cfabort>
</cfif>
The way you're coupling your business logic to your display leaves a bit to be desired. You could probably benefit from reading up on MVC and separation of concerns.
From the perspective of your logic, your validation rules seem fine, but you're doing the validation twice, which seems excessive: each element, then all elements. This is in part due to the problem I highlight above.
I'd give some thought to stop thinking procedurally, and think in a more OO fashion, and define the notion of a User.cfc, and have some sort of validation service (see ValidateThis). Or something like that.
Lastly, this is not really the sort of question best asked on Stack Overflow, but would be good for Code Review. There's no one answer for this question, so people will be inclined to suggest closing it for being "primarily opinion-based".
I'm also gonna retag this as just "ColdFusion" rather than "ColdFusion 10", as it really has nothing specifically to do with CF10, it's just a CFML question. You'll get a bigger audience with it marked as just "ColdFusion".
Instead of sharing code with you, I would like to introduce the concepts to you. The first thing you should do is read the OWASP recommendations for Data Validation. In it they suggest that there are four strategies for validating data, and they should be used in the following order. I will post some excerpts here but I strongly recommend you read the entire article.
Accept known good
This strategy is also known as "whitelist" or "positive" validation. The idea is that you should check that the data is one of a set of tightly constrained known good values. Any data that doesn't match should be rejected.
Reject known bad
This strategy, also known as "negative" or "blacklist" validation is a weak alternative to positive validation. Essentially, if you don't expect to see characters such as %3f or JavaScript or similar, reject strings containing them. This is a dangerous strategy, because the set of possible bad data is potentially infinite. Adopting this strategy means that you will have to maintain the list of "known bad" characters and patterns forever, and you will by definition have incomplete protection.
Sanitize
Rather than accept or reject input, another option is to change the user input into an acceptable format
No validation
This is inherently unsafe and strongly discouraged. The business must sign off each and every example of no validation as the lack of validation usually leads to direct obviation of application, host and network security controls.
The article goes on to discuss each of these in greater detail and much more.
This is another way. You can decide for yourself whether or not it is better.
Step 1 - create an error message variable.
<cfset ErrorMessage = "">
Step 2 - Do your checks. If you see something you don't like, append text to your variable.
<cfif len(trim(form.LoginName)) gt 0>
<cfset ErrorMessage &= "<h3>Login Name is required</h3>">
</cfif>
more checks
Step 3 - Check the length of your error message variable
<cfif len(ErrorMessage) gt 0>
display it
<cfelse>
code for no errors
</cfif>
In addition to all this, you probably want to check to see if the page request actually came from your form page. You can use cgi.http_referrer for that.
One more thing. Instead of an anchor tag back to the form page like this,
<p>Return to users list.</p>
You can use javascript so that page does not have to reload in the browser.
<p>Return to users list.</p>
In our index files, at the top of the file we will generally <cfparam> out the variables that will come from the URL, form or wherever. However, we are getting a lot of bots hitting us with things like www.example.com/survey/index.cfm?nPageNumber=-1 meaning a cfparam like this:
<cfparam name="request.parameters.nPageNumber" default="1" type="numeric" />
will fail due to the nonsense a bot is putting into the querystring.
I find myself increasingly having to write my cfparams like this:
<cfif structKeyExists(request.parameters,"nPageNumber") AND isNumeric(request.parameters.nPageNumber)>
<cfparam name="request.parameters.nPageNumber" default="1" type="numeric" />
<cfelse>
<cfset request.parameters.nPageNumber = 1>
</cfif>
While this solves the issue, I can't help but feel this solution is not the best/most efficient. Am I using cfparam correctly or are there better ways of doing this?
Ensuring the existence of a variable, and validating its value are two separate tasks.
In the case of URL and Form your existence code should be something like:
<cfparam name="URL.nPageNumber" default="1" type="string">
The use of type there is just to ensure that nothing truly bizarre has happened such as the value is a struct or query, etc. You don't want to get specific at this point since you want a graceful error and not at 500 for the user.
Once you've made sure that the value exists, you then need to validate the value.
<cfif isNumeric(URL.nPageNumber) EQ false OR URL.nPageNumber LT 1 OR URL.nPageNumber GT Variables.MaxPages>
<cfset ArrayAppend(Variables.ErrorArray, "Incorrect page number requested.")>
</cfif>
You could force the value to something sane, but see Martian Headsets for a counter argument to the robustness principle.
Providing an error message instead of "displaying something sane" will inform your users that they are doing something wrong, and means you're not forced to use canonical urls if you're not already (although you should be).
Yes, it's more work. You can devise some abstraction for it all, but at the raw level, that's what you should be doing with your cfparams and validation.
In the situation where you don't need to have a friendly response, such as a bot or a request that is obviously a hack or probing, there's the additional option to serve a "400" response code. The w3c defines the response as "The request had bad syntax or was inherently impossible to be satisfied." here and "The request could not be understood by the server due to malformed syntax. The client SHOULD NOT repeat the request without modifications." here.
I'm having issues with ColdFusion's heap. Here's a little example application I'm experimenting with. I thought after cfinvoke calling the init method it destroys all variables local to the component. But apparently it's not the case. The application works as it is below but if I add a zero to the loop in index.cfm it breaks. What's stored in the heap to cause that? Is there a way around this?
index.cfm:
<cfloop from="1" to="1000" index="i">
<cfinvoke component="test" method="init" returnvariable="x">
</cfloop>
<cfoutput><p>#x#</p></cfoutput>
test.cfc:
<cfcomponent output="false">
<cffunction name="init" returntype="string">
<cfset var test = structNew()>
<cfloop from="1" to="1000" index="i">
<cfset test[i] = i>
</cfloop>
<cfreturn Now()>
</cffunction>
</cfcomponent>
And here's the error message:
SEVERE: Servlet.service() for servlet CfmServlet threw exception
javax.servlet.ServletException: ROOT CAUSE:
java.lang.OutOfMemoryError: Java heap space`
Any help would be appreciated.
This is a known problem and CURRENTLY there is no way around it. It seems that CF DOES NOT handle garbage collection properly. As far as I can tell it will not destroy and free up memory used by component until after the request has finished.
Because of that, depending on how many properties your components has will determine how many of them you can create until you get the heap error.
I know this doesn't help your problem, but at least now you can take measures to prevent it.
rio747 is correct, the memory does not free until the request has ended, so you are simply running out of memory. I would venture to say you can probably run it and get the error, remove the 0 and it will run; the java garbage collection is running, but not until the request is ended.
Your only options for resolution is to not create as many objects in a single request (maybe some more information about why this is necessary?) or to increase the amount of memory available to coldfusion.
I don't think this is really a problem or CF not handling garbage collection properly; from a java standpoint as long as that request is active the object could still be used and referenced. I mean, you are right about the cfinvoke and how it works, but that doesn't mean that the jvm can actually run garbage collection on those memory locations until the request is ended.
I have just posted a new blog entry which shows how you can do programmatic garbage collection if your memory gets below a certain threshold.. Hope it helps:-
[http://www.beetrootstreet.com/blog/index.cfm/2009/6/25/Clearing-ColdFusion-memory-using-garbage-collection-when-memory-gets-low][1]
Been going over my predecessor's code and see usage of the "request" scope frequently. What is the appropriate usage of this scope?
There are several scopes that are available to any portion of your code: Session, Client, Cookie, Application, and Request. Some are inadvisable to use in certain ways (i.e. using Request or Application scope inside your Custom Tags or CFC's; this is coupling, violates encapsulation principles, and is considered a bad practice), and some have special purposes: Cookie is persisted on the client machine as physical cookies, and Session scoped variables are user-specific and expire with the user's session on the website.
If a variable is extremely unlikely to change (constant for all intents and purposes) and can simply be initialized on application startup and never written again, generally you should put it into Application scope because this persists it between every user and every session. When properly implemented it is written once and read N times.
A proper implementation of Application variables in Application.cfm might look like this:
<cfif not structKeyExists(application, "dsn")>
<cflock scope="application" type="exclusive" timeout="30">
<cfif not structKeyExists(application, "dsn")>
<cfset application.dsn = "MyDSN" />
<cfset foo = "bar" />
<cfset x = 5 />
</cfif>
</cflock>
</cfif>
Note that the existence of the variable in the application scope is checked before and after the lock, so that if two users create a race condition at application startup, only one of them will end up setting the application variables.
The benefit of this approach is that it won't constantly refresh these stored variables on every request, wasting the user's time and the server's processing cycles. The trade-off is that it is a little verbose and complex.
This was greatly simplified with the addition of Application.cfc. Now, you can specify which variables are created on application startup and don't have to worry about locking and checking for existence and all of that fun stuff:
<cfcomponent>
<cfset this.name = "myApplicationName" />
<cffunction name="onApplicationStart" returnType="boolean" output="false">
<cfset application.dsn = "MyDSN" />
<cfset foo = "bar" />
<cfset x = 5 />
<cfreturn true />
</cffunction>
</cfcomponent>
For more information on Application.cfc including all of the various special functions available and every little detail about what and how to use it, I recommend this post on Raymond Camden's blog.
To summarize, request scope is available everywhere in your code, but that doesn't necessarily make it "right" to use it everywhere. Chances are that your predecessor was using it to break encapsulation, and that can be cumbersome to refactor out. You may be best off leaving it as-is, but understanding which scope is the best tool for the job will definitely make your future code better.
This is a very subjective question, and some would even argue that it is never "appropriate" to use the request scope in modern ColdFusion applications.
With that disclaimer out of the way, let's define what the request scope is and where it would be useful.
The request scope is the absolute global scope in a single ColdFusion page request. It is not a shared scope, like application, server, client, and session scopes, so locking is not necessary to make it threadsafe (unless you spawn worker threads from a single request using CF8's CFTHREAD tag). As a global scope, it is a very convenient way to persist variables through any level in the request's stack without having to pass them from parent to caller. This was a very common way to persist variables through nested or recursive Custom Tags in older CF apps.
Note that while many applications use this scope to store application-level variables (configuration settings, for example), the big (and sometimes subtle) difference between the request scope and the application scope is that the value of the same request-scoped variable can differ between individual page requests.
I would guess that your predecessor used this scope as a means to conveniently set variables that needed to survive the jump between encapsulated or nested units of code without having to pass them around explicitly.
Okay, I just wanted to comment on your code. Please forgive me if I seem crazy. But you already verified that the structKeyExists in the beginning. Since you know it's going to be true, it wouldn't make sense to run another check. So my version of it would be this... But thats just me.
<cfif not structKeyExists(application, "dsn")>
<cflock scope="application" type="exclusive" timeout="30">
<cfset application.dsn = "MyDSN" />
<cfset foo = "bar" />
<cfset x = 5 />
</cflock>
</cfif>
Alright.
I've been writing my company's framework that will be used to power our site.
I use the request variable to set certain data that would be available to the other CFC's I had to do this so the data would be available throughout the application, without the need to continually pass in the data. I honestly believe that using request , and application as long as its a static function component then you should not have a problem. I'm not sure if I am wrong in my thinking with this but once I release the framework we will see.