it is very nice solution, one thing that i would change is that you have to call that magic class MethodTimeLogger exactly same way, I would prefer some way of registering class that implements some interface with the package
Another strategy would be to create middleware that is added to the very beginning of the pipeline on the way in… record the time, perhaps even store it in the request object…. Call Next()… then record the time to get your elapsed time for the entire request. In addition, one of the strategies we did is fire off a Kafka message (event hub / service bus works) over time u have some excellent telemetry that can be monitored.! Then efficiently send an entire batch of request telemetry to a Kafka topic every second And finally, instead of making a network call for every request, we would buffer up a batch of messages…
Amazing, thank you Nick and Simon, this is the easy mode benchmarking I needed to actually start profiling in depth. is this compiled out of release builds?
Very interesting. Symfony uses such data to show in their debug toolbar/panel/graph how much time each part of the framework takes. You can even group stuff together like all repositories and see how much time your db interaction takes, or your template compilation or whatever
Why can't this be done with a Source Code Generator? Having to include Fody for IL weaving seems like both a heavy handed and an old fashioned approach. It is expected that all developers using Fody become a Patron on OpenCollective.
Great video. Yesterday I was searching for middleware to measure the performance of the endpoints but now I see it can be done quite easy with this package too
Hey @Nick Chapsas, cool utility! Love how you're always sharing these neat little libraries I never would have heard of otherwise! One important question that comes to mind with the way this is implemented: what happens to the stack trace from methods that have this attribute? Will they still have the original method's name and class?
Nice, but currently, you can't update Git from Visual Studio while this is installed - had to do a restart and remove it to commit. I am sure it will be fixed in time and will be most useful to use.
I love Aspects. I've used PostSharp for many years, which I believe, uses Fody under the covers - although PostSharp is not free (commercially anyway). This is a great way of removing those cross cutting concerns from your logic code. One aspect I've used is for tracing stored procedure parameters with a single method attribute.
I'm not suggest any IL Weavering type libraries at all, because it it has some unwnated side effects like mismatched stacktrace compare to actual production code and so on....
I don't know how this could be mixed with OpenTelemetry spans, because the way your log method just gets the elapsed time means that even if you make your logger build the spans manually, and you set the span's start/end times, you would still lose the parent child relationship with your spans. You could definitely implement metrics, but the spans seem dubious. if the package let you run your own method to start the stopwatch in addition to the logging, then you could do spans. The other weavers i've looked at like MethodBoundaryAspect and MethodDecorator have issues weaving correctly into the IL of the async/await state machine. MethodBoundaryAspect you can do a task.continueWith, but that still screws up the parent/child relationship.
I wonder how would you write those values in a time series database if your applicacion has not too many users at the same time of course. To observe the behaviour over time of the application. To observe endpoints.
I don't quite understand the last part - how do you use the non-allocating way of time measurement with Fody still, since the use of StopWatch is completely controlled by Fody itself?
I never measure performance unless there the code is slow, and not even then measure the code, I just look at the code and change it for better performance
They let you choose which issues rider highlights (only up to errors, only up to warnings, etc.) and control other visual hints like enabling/disabling inline information.
How does this compare to environments using interceptors? It felt pretty similar just without marking up the code with details about cross-cutting concerns. Just setting it up when you are defining your dependency map and that's it. I've been happy with it so far.
Very cool post! I have a question, please. You showed how to send a variable in the message, but is it possible to send an object property? For example this is my signature method: ```public async Task GetAllAppointments([FromQuery] GetAllAppointmentsQuery paginationQuery)``` I want to send a property from the object like this: ```[Time("Retrieved {paginationQuery.PageSize} appointments.")]``` This gives an error, and from what I read in the official docs of the library, it doesn't support referencing properties. Is there a workaround?
Hi @nickchapsas, are you still planning to make video how to join this nugget with some telemetry? Also I would like to know if is possible to get hierarchical view of call when method call another methods.
Hmm, but how it worked ? How the code was changed based on an Attribute before the compilation ? I've checked IL, and there is a Stopwatch already injected. I think this is a big topic to explain, but advanced...
Is there a posibility to have an instance of the MethodTimeLog and not a static one? I wanted to log all method in one log in every request. But since since the Log function is static, I can't do it. Is there any better way than this?
I've made my own class to get this, but this is another level ... thanks a lot. Another that I would like to have, is a Translation method that can also works with variables in between the text, any ideas?
What you need is an i18n library (internationalization). How they work is you create what is basically a dictionary of keys (identifiers for phrases, e.g. "login.error.invalidPassword"), and then you specify a value which is specific to each language, optionally passing parameters to fill in the places where variables appear as you mentioned. I haven't used one for C# before, personally, I do this on the front end instead where all the language text is stored in a json file. From a quick search on github for a C# library though, if it was me I would maybe try TypealizR - from the readme, it looks like it works in the way I would expect, and was last updated yesterday, so means it's still maintained. But again I have not used it myself as I don't do this on the backend.
While Fody is a cool project they expect you to pay money for it, seems a bit counter intuitive for a open source project... so i don't know if i would use the project unless i must...
is it possible to automatically log all the input parameters without having to specify their names in the TimeAttribute ctor? Or even take complex object and serialize them to json
What looks like an !. operator made me double take. Is that null forgiving, but thats for method calls? Or are you using a pre release version of the compiler?
@@nickchapsas Ah yes, they added all the warnings on reference intent. While I am on C# 10 I disabled that functionality as the code originates from dotnet rc1 and I don't have the time to fix so many warnings. Probably make it more reliable when I do though.
Nick how do you find these small libraries? Maybe make a video on that topic? If you put in a solid amount of time, do you think it's worth for "non creator developers" to do the same or should we just wait for the video/article? I am pretty involved and up to date. I follow people on youtube, I check conference talks, I am checking newsletters/aggregators, I randomly look for articles and videos like "10 useful nuget packages". But I feel I could still do with way more "hey this exists, this is probably useful some time in the future" type of knowledge.
In regard to telemetry, I'd love to see an OpenTelemetry video. Aside from logs, both the traces and metrics SDKs are stable.
It is definitely coming
Cool password bro
@@nickchapsasYes pleeease! Would love an Open Telemetry video
it is very nice solution, one thing that i would change is that you have to call that magic class MethodTimeLogger exactly same way, I would prefer some way of registering class that implements some interface with the package
It is a great library. I was always waiting for something like this to measure time, easy, clean, and beautiful. Thanks Nick!!!!!
Another strategy would be to create middleware that is added to the very beginning of the pipeline on the way in… record the time, perhaps even store it in the request object…. Call Next()… then record the time to get your elapsed time for the entire request. In addition, one of the strategies we did is fire off a Kafka message (event hub / service bus works) over time u have some excellent telemetry that can be monitored.! Then efficiently send an entire batch of request telemetry to a Kafka topic every second
And finally, instead of making a network call for every request, we would buffer up a batch of messages…
I do the same with NServicebus. Really effective way to monitor queue times.
You said it first... :)
Cheers Nick. This is super useful. We have stop watches everywhere and its kinda clunky. Definitely going to investigate this.
Thank you, this helps so much when working on new projects!
I fucking love both you and this nugget provider for making and showcasing such a useful tool!
Wow thanks, that's exactly what i needed right now!
There's also Serilog Timings library; writes measurements to log.
It does but it is Serilog specific. You can plug any logging provider or metrics collector in this one which I prefer
Amazing, thank you Nick and Simon, this is the easy mode benchmarking I needed to actually start profiling in depth. is this compiled out of release builds?
Great video as always, thank you for the hard work.
Thanks! It is a really nice library (and underrated, only several hundreds stars) :)
Great video Nick. Ive been looking for a better way to measure methods for a long time. Thanks for sharing!
Thanks for the free course and awesome work
Hey Nick, nice video. Aside from time performance, can I log the allocated memory by the method?
Very interesting. Symfony uses such data to show in their debug toolbar/panel/graph how much time each part of the framework takes. You can even group stuff together like all repositories and see how much time your db interaction takes, or your template compilation or whatever
Content starts at 2:15
Why can't this be done with a Source Code Generator? Having to include Fody for IL weaving seems like both a heavy handed and an old fashioned approach.
It is expected that all developers using Fody become a Patron on OpenCollective.
Great video. Yesterday I was searching for middleware to measure the performance of the endpoints but now I see it can be done quite easy with this package too
Hey @Nick Chapsas, cool utility! Love how you're always sharing these neat little libraries I never would have heard of otherwise!
One important question that comes to mind with the way this is implemented: what happens to the stack trace from methods that have this attribute? Will they still have the original method's name and class?
Yes! Basically all that code will be as if it was in the main method at compile time
Nice, but currently, you can't update Git from Visual Studio while this is installed - had to do a restart and remove it to commit. I am sure it will be fixed in time and will be most useful to use.
Hi Nick, Thank you for the great video ,
How does the Package work or How Can I make method attribute work as decorator ?
I love Aspects. I've used PostSharp for many years, which I believe, uses Fody under the covers - although PostSharp is not free (commercially anyway). This is a great way of removing those cross cutting concerns from your logic code. One aspect I've used is for tracing stored procedure parameters with a single method attribute.
Fantastic! Thanks!
good explaination
I'm not suggest any IL Weavering type libraries at all, because it it has some unwnated side effects like mismatched stacktrace compare to actual production code and so on....
agree, especially if it create exception trap
I don't know how this could be mixed with OpenTelemetry spans, because the way your log method just gets the elapsed time means that even if you make your logger build the spans manually, and you set the span's start/end times, you would still lose the parent child relationship with your spans.
You could definitely implement metrics, but the spans seem dubious.
if the package let you run your own method to start the stopwatch in addition to the logging, then you could do spans.
The other weavers i've looked at like MethodBoundaryAspect and MethodDecorator have issues weaving correctly into the IL of the async/await state machine.
MethodBoundaryAspect you can do a task.continueWith, but that still screws up the parent/child relationship.
Please do an Azure course.
Only if Azure sponsors it
@@nickchapsas they'd be stupid not to tbh
@@nickchapsas you should ask them to sponsor it, I'll love to see an Azure course!
I wonder how would you write those values in a time series database if your applicacion has not too many users at the same time of course. To observe the behaviour over time of the application. To observe endpoints.
Is it compatible with AOT?
I don't quite understand the last part - how do you use the non-allocating way of time measurement with Fody still, since the use of StopWatch is completely controlled by Fody itself?
Hmmm cool stuff. Does AOP libraries use assembly weaving too to handle cross-cutting concerns?
How did you go to the code that is converted to ??
I never measure performance unless there the code is slow, and not even then measure the code, I just look at the code and change it for better performance
Very good
I have been watching your content a long time, i have never used Raider it seems nice but what do the pencils/fence in the lower right corner do?
They let you choose which issues rider highlights (only up to errors, only up to warnings, etc.) and control other visual hints like enabling/disabling inline information.
@@LeMustache thanks!
How does this compare to environments using interceptors? It felt pretty similar just without marking up the code with details about cross-cutting concerns. Just setting it up when you are defining your dependency map and that's it. I've been happy with it so far.
Hi Nick !
this is cool!
Very cool post!
I have a question, please. You showed how to send a variable in the message, but is it possible to send an object property? For example this is my signature method:
```public async Task GetAllAppointments([FromQuery] GetAllAppointmentsQuery paginationQuery)```
I want to send a property from the object like this: ```[Time("Retrieved {paginationQuery.PageSize} appointments.")]```
This gives an error, and from what I read in the official docs of the library, it doesn't support referencing properties. Is there a workaround?
Hi @nickchapsas, are you still planning to make video how to join this nugget with some telemetry? Also I would like to know if is possible to get hierarchical view of call when method call another methods.
Hmm, but how it worked ?
How the code was changed based on an Attribute before the compilation ?
I've checked IL, and there is a Stopwatch already injected. I think this is a big topic to explain, but advanced...
Is there a posibility to have an instance of the MethodTimeLog and not a static one? I wanted to log all method in one log in every request. But since since the Log function is static, I can't do it. Is there any better way than this?
does this work with a method that yields results into an IEnumerable output?
I have actually never tried that. I'd have to give it a go and see
I've made my own class to get this, but this is another level ... thanks a lot.
Another that I would like to have, is a Translation method that can also works with variables in between the text, any ideas?
What you need is an i18n library (internationalization). How they work is you create what is basically a dictionary of keys (identifiers for phrases, e.g. "login.error.invalidPassword"), and then you specify a value which is specific to each language, optionally passing parameters to fill in the places where variables appear as you mentioned. I haven't used one for C# before, personally, I do this on the front end instead where all the language text is stored in a json file. From a quick search on github for a C# library though, if it was me I would maybe try TypealizR - from the readme, it looks like it works in the way I would expect, and was last updated yesterday, so means it's still maintained. But again I have not used it myself as I don't do this on the backend.
Does grabbing and dealing with MethodIndo has any performance issues when using this library?
The link in the description is no longer valid!
I think that datadog is doing the same things for you already out of the box
Does this work with aot?
Can't signup on the website as a new User, there is a problem with teachable
While Fody is a cool project they expect you to pay money for it, seems a bit counter intuitive for a open source project... so i don't know if i would use the project unless i must...
unreal engine is also open-source, but they still expect you to pay for it. it's just a business-model.
Found the "use open source but never ever support the developers behind it" guy
Found the "use open source but never ever support the developers behind it" guy
Open source != free.
Open source is not necessary free
can you export them to a dashboard in aws? because using app.metrics its very easy since in expose an api endpoint for metrics
Sure it is. You can plug in App Metrics, OpenTelemetry or whatever you want
I really want to buy your course, however £80.00 is too pricey when you live in SA
is it possible to automatically log all the input parameters without having to specify their names in the TimeAttribute ctor? Or even take complex object and serialize them to json
Yeah absolutely. You can do that by accessing them in the MethodInfo object. They are just an array that you can loop
@@nickchapsas i thought it was possible for params metadata, but can you even get the runtime value of them?
What looks like an !. operator made me double take. Is that null forgiving, but thats for method calls? Or are you using a pre release version of the compiler?
It’s just suppressing the null warning. It’s been there since C# 8 I’m pretty sure
@@nickchapsas Ah yes, they added all the warnings on reference intent. While I am on C# 10 I disabled that functionality as the code originates from dotnet rc1 and I don't have the time to fix so many warnings. Probably make it more reliable when I do though.
oh yeahh
Damn just bought 2 courses and can't use the discount now
like if your playback settings are 0.5x to understand him
Could have done with this two weeks ago. 🙄
First
I just got rid of Fody in my projects.
Code generation has to be better than that.
Fody is assembly weaving not code generation
@@nickchapsas You're right. But Fody interferes with build. It's super annoying. So i don't want it
Third!
The C# team should add python-like decorators
This is not the video i have searched for, Google
This was 9 lines. disappointing.
First comment
Datadog now support OpenTelelerty. The way more powerful combination
public class StopWatch : System.Diagnostics.Stopwatch
{
public StopWatch() : base(){}
public TimeSpan CalculateDuration(Action a)
{
base.Reset();base.Start();a();base.Start();return base.Elapsed;
}
public (TimeSpan Duration, TResult Result) CalculateDuration(Func function)
{
TResult result=default;
base.Reset();base.Start();result=function();base.Start();return (Duration: base.Elapsed, Result:result);
}
public (TimeSpan Duration, TResult Result) CalculateDuration(Func asyncFunction)
{
TResult result=default;
base.Reset();base.Start();base.Start();Call().RunSynchronously();return (Duration: base.Elapsed, Result:result);
async Task Call()
{
result=await asyncFunction();
}
}
}
Would it just be simpler to add the Prometheus library and now you have all that info on /metrics ?
Fody isn't free, it requires a mandatory donation matching the number of devs using the code base or something like that.
i have notice in the end of video you said better to use telemetry, then why not let us know how to if that is the best practice
How much does Fody cost?
Love these types of videos Nick!
Nick how do you find these small libraries? Maybe make a video on that topic? If you put in a solid amount of time, do you think it's worth for "non creator developers" to do the same or should we just wait for the video/article?
I am pretty involved and up to date. I follow people on youtube, I check conference talks, I am checking newsletters/aggregators, I randomly look for articles and videos like "10 useful nuget packages".
But I feel I could still do with way more "hey this exists, this is probably useful some time in the future" type of knowledge.