Issue 140 #142

bgreen1005 · 2020-02-21T12:39:00Z

140: This commit fixes issue 140 (https://github.com/jenkinsci/code-coverage-api-plugin/issues/140) by adding back in the code that iterates through the file path and finds a file that matches. I have also optimized the lookup by getting all of the file paths in one go. This drastically improves performance for jacoco. One thing to note, I've added some code to log duplicates. This is due to the fact that Jacoco only reports the class name, not the directory and so duplicates are possible.

I noticed a massive performance improvement with these changes. Generating our code coverage report has dropped from about 8 minutes to 5 or 6 seconds.

I faced a few issues with the POM file, and had to update some dependencies. I wasn't sure if that was just a local problem, so haven't checked in the changes but I can do if need be.

I've also added a few fixmes due to unnecessary code in DefaultSourceFileResolver. Happy to fix if you'd like :)

…ates through the file path and finds a file that matches.

cizezsy

Thanks for your fix, I have left some comments :P

cizezsy · 2020-02-24T08:45:38Z

src/main/java/io/jenkins/plugins/coverage/source/DefaultSourceFileResolver.java

+    private Map<String, FilePath> createSourceFileMapping(FilePath workspace, TaskListener listener) {
+        try {
+            return Arrays
+                    .stream(workspace.list("**/*.java"))


This plugin also supports other programming languages like JavaScript, C#, C++, so this might not work for these languages.

Maybe we can use another optimization method here. We used FilePath(workspace).list("**/" + sourceFilePath) to search source file for each source file before, it is really slow and does a lot of repeated work.

But, most of the source files have the same parent directory, we can use the wildcard to search the source file if we are unable to find it directly under the workspace and possiblePaths. Once we successfully find the source file by using the wildcard, we will add its parent path to possiblePaths so that we can reuse it to avoid search next time.

What do you think of this change? Thanks~

@cizezsy Ahh sorry, I totally mean to change that from **/*.java to **/*, but i'm happy to make the changes you've suggested if you think that's preferable? Just to verify what it is you're looking for there: in tryFindSourceFile, instead of looking at the wildcard every time, we check to find the file in our list of possible paths. If we can't find it there, we load it with the wildcard and add the parent path to the list of possible paths?

Does doing that also resolve your other comment about doing the calculating in the agent instead of master? I guess it does?

It feels like that wouldn't give us anywhere near the performance benefits that the above change does (assuming I change **/*.java to **/*. What are your thoughts on that?

Just to verify what it is you're looking for there: in tryFindSourceFile, instead of looking at the wildcard every time, we check to find the file in our list of possible paths. If we can't find it there, we load it with the wildcard and add the parent path to the list of possible paths?

Yes.

Does doing that also resolve your other comment about doing the calculating in the agent instead of master? I guess it does?

Yes :)

It feels like that wouldn't give us anywhere near the performance benefits that the above change does (assuming I change **/.java to **/. What are your thoughts on that?

Yeah, you are right, that will not increase the performance compare with your changes. But, in your implementation, you used the file name as the key of sourceFileMapping. And sourceFilePath might not be a file name instead of a path of the source file(like Cobertura). You won't get a correct source file form the map in this situation.

@cizezsy This change should have no impact on Corbertura at all. Previously at the end of tryFindSourceFile, we were just returning null. That will be exactly the same if the source file mapping isn't found in our map. Also, tryFindSourceFile has lots of code to handle the case where the source file is absolute, so Corbertura shouldn't need the fallback anyway I don't think. We have about 7000 classes, and this change has improved performance from about 10 minutes to 5 seconds. We'd really like it if we could keep those performance benefits 🤞

@bgreen1005 Sounds reasonable to me :p

I will test the changes on my computer. If there are no other problems, we can get this merged. Thanks~

@cizezsy Perfect! Thank you 🙂 Hopefully no issues 🤞

@cizezsy How's it looking? Have you had chances to check this? 🙂

@bgreen1005 I will try it this weekend. Sorry for the delay, I am very busy this week :(

@cizezsy no problem 🙂 Thanks for the heads up.

cizezsy · 2020-02-24T08:47:34Z

src/main/java/io/jenkins/plugins/coverage/source/DefaultSourceFileResolver.java

@@ -78,8 +80,10 @@ public void resolveSourceFiles(Run<?, ?> run, FilePath workspace, TaskListener l

        listener.getLogger().printf("%d source files need to be copied.%n", paints.size());

+        final Map<String, FilePath> sourceFileMapping = createSourceFileMapping(workspace, listener);


It is better to do calculating in the agent to reduce the Jenkins master workload.

@cizezsy Sorry for the stupid question, but would you mind explaining how you can tell which of this is run on master and which is run on agents?

cizezsy

LGTM, thanks for your contribution~

(I think we need another PR to improve the case which two file have the same name, but we can get this merged now :P

bgreen1005 · 2020-03-01T16:26:44Z

Great thanks! Yeah I agree. We could do with being able to differentiate between potential duplicates. How often do you release?

cizezsy · 2020-03-09T10:14:16Z

@bgreen1005 I will perform a release this week :)

medianick · 2020-04-20T18:03:39Z

src/main/java/io/jenkins/plugins/coverage/source/DefaultSourceFileResolver.java

+    private Map<String, FilePath> createSourceFileMapping(FilePath workspace, TaskListener listener) {
+        try {
+            return Arrays
+                    .stream(workspace.list("**/*"))


Why list all files in the workspace (which would include compiled binaries, node_modules, etc.) rather than only those referenced by the coverage report(s) being processed? See #150 for something that seems to be caused by this change.

If you look at this more closely, you will see that it is executed if the source file we are looking for was not found in the usual way.

And this code resolves source files not only for Cobertura, but also for formats like Jacoco, which does not have such a direct path to the source files.

I'll remove the warnings to don't disturb people

Yeah, as a fallback mechanism specifically -- and one that seems to be deliberately using just the filenames without paths -- it doesn't seem worth the noise to have all these duplicates get logged. Otherwise you're logging warnings for a map that might not even get used (and ideally would not, if all went well).

I'm looking in the code and don't see this as a fallback, but directly called. Am I missing something.
I'm investigating why this plugin is taking over 6 hours in CoverageProcessor.convertToResults, while the Cobertura plugin runs in only 21 seconds. ( I know it's this part of the code per the log out time start of "Publishing Coverage report..." vs the next output of "A total of X reports were found".

Yes, that mapping is always constructed -- which may be the cause of your problem -- but is only used if it cannot find a match from the file names/paths in the coverage reports. I believe that's why it was being described as a "fallback".

140: This commit fixes issue 140 by adding back in the code that iter…

4081720

…ates through the file path and finds a file that matches.

bgreen1005 changed the title ~~140: This commit fixes issue 140: https://github.com/jenkinsci/code-coverage-api-plugin/issues/140~~ Issue 140 Feb 21, 2020

bgreen1005 requested a review from cizezsy February 21, 2020 12:41

cizezsy suggested changes Feb 24, 2020

View reviewed changes

140: Now matching all files instead of just java files

196f8ce

cizezsy approved these changes Mar 1, 2020

View reviewed changes

cizezsy merged commit a29a9d5 into jenkinsci:dev Mar 1, 2020

medianick reviewed Apr 20, 2020

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Issue 140 #142

Issue 140 #142

bgreen1005 commented Feb 21, 2020 •

edited

Loading

cizezsy left a comment

cizezsy Feb 24, 2020

bgreen1005 Feb 24, 2020 •

edited

Loading

cizezsy Feb 24, 2020 •

edited

Loading

bgreen1005 Feb 24, 2020

cizezsy Feb 25, 2020

bgreen1005 Feb 25, 2020

bgreen1005 Feb 26, 2020

cizezsy Feb 28, 2020

bgreen1005 Feb 28, 2020

cizezsy Feb 24, 2020

bgreen1005 Feb 24, 2020 •

edited

Loading

cizezsy left a comment

bgreen1005 commented Mar 1, 2020

cizezsy commented Mar 9, 2020

medianick Apr 20, 2020

tempora-mutantur Apr 20, 2020

medianick Apr 20, 2020

trash-80 May 18, 2020

medianick May 20, 2020

		@@ -78,8 +80,10 @@ public void resolveSourceFiles(Run<?, ?> run, FilePath workspace, TaskListener l

		listener.getLogger().printf("%d source files need to be copied.%n", paints.size());

		final Map<String, FilePath> sourceFileMapping = createSourceFileMapping(workspace, listener);

Issue 140 #142

Issue 140 #142

Conversation

bgreen1005 commented Feb 21, 2020 • edited Loading

cizezsy left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bgreen1005 Feb 24, 2020 • edited Loading

Choose a reason for hiding this comment

cizezsy Feb 24, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bgreen1005 Feb 24, 2020 • edited Loading

Choose a reason for hiding this comment

cizezsy left a comment

Choose a reason for hiding this comment

bgreen1005 commented Mar 1, 2020

cizezsy commented Mar 9, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bgreen1005 commented Feb 21, 2020 •

edited

Loading

bgreen1005 Feb 24, 2020 •

edited

Loading

cizezsy Feb 24, 2020 •

edited

Loading

bgreen1005 Feb 24, 2020 •

edited

Loading