{"user_signed_in":false,"allow_comments":true,"is_author":false,"request":{"endpoint":"https://exercism.org/api/v2/tracks/go/exercises/paasio/community_solutions/zeimusu/comments","options":{"initial_data":{"items":[{"uuid":"3af61acc-08d3-11ec-81c5-853579c25f94","author":{"handle":"zeimusu","avatar_url":"https://assets.exercism.org/avatars/122383/0","flair":null,"reputation":"294"},"content_markdown":"@eraserix commented: I guess you (and many others) open yourself up for a race condition if you don't lock the mutex in {Read,Write}Count(). ... I think you are exactly right, and while I can't be bothered right now to correct this you might think about how this could be tested. After all, this Passes all tests. Is it possible to construct a test that illustrates that a race condition exists?","content_html":"

@eraserix commented:

I guess you (and many others) open yourself up for a race condition if you don't lock the mutex in {Read,Write}Count(). ...

I think you are exactly right, and while I can't be bothered right now to correct this you might think about how this could be tested. After all, this Passes all tests. Is it possible to construct a test that illustrates that a race condition exists?

","updated_at":"2018-07-13T10:01:29Z","links":{}},{"uuid":"3af61bd0-08d3-11ec-81c5-853579c25f94","author":{"handle":"eraserix","avatar_url":"https://assets.exercism.org/avatars/175619/0","flair":null,"reputation":"178"},"content_markdown":"@zeimusu commented: I think you are exactly right, and while I can't be bothered right now to correct this you might think about how this could be tested. After all, this Passes all tests. Is it possible to construct a test that illustrates that a race condition exists? I just created a pull-request with new testcases that check for this particular race condition.","content_html":"

@zeimusu commented:

I just created a pull-request with new testcases that check for this particular race condition.

","updated_at":"2018-07-13T10:01:29Z","links":{}},{"uuid":"3af61c98-08d3-11ec-81c5-853579c25f94","author":{"handle":"zeimusu","avatar_url":"https://assets.exercism.org/avatars/122383/0","flair":null,"reputation":"294"},"content_markdown":"WILT This took ages for me to get my head around. The first thing: You need to create a struct to implement the ReadCounter and WriteCounter interfaces. This struct needs to have functions as part of it. But the same struct will do for both interfaces. More obviously it needs a way of recording the number of calls and the number of bytes transferred, and a locking mechanism Then to implement the interaces you need to write Read, Write, ReadCount and WriteCount methods. The Read and Write methods need to call their read or write function, and then update the number of calls and bytes. Now it might seem natural to use a pointer receiver, but that is not allowed by the interface. So the the calls and bytes fields need to be pointers, so they can be updated by the Read method. Lock both Read and Write, or the tests don't pass (it should be possible to do Reads asynchonously, but it seems that some get lost if you try that) Perhaps you only need to lock the section that changes the numbers of calls and bytes. Now for the initialisation you need to set up calls with a the address of a variable with value zero. There is probably a better way to do that, but this passes. There is rather a lot of copy and paste here. Probably many improvements could be made.","content_html":"

WILT

This took ages for me to get my head around.

The first thing: You need to create a struct to implement the ReadCounter and WriteCounter interfaces. This struct needs to have functions as part of it. But the same struct will do for both interfaces. More obviously it needs a way of recording the number of calls and the number of bytes transferred, and a locking mechanism

Then to implement the interaces you need to write Read, Write, ReadCount and WriteCount methods.

The Read and Write methods need to call their read or write function, and then update the number of calls and bytes. Now it might seem natural to use a pointer receiver, but that is not allowed by the interface. So the the calls and bytes fields need to be pointers, so they can be updated by the Read method.

Lock both Read and Write, or the tests don't pass (it should be possible to do Reads asynchonously, but it seems that some get lost if you try that) Perhaps you only need to lock the section that changes the numbers of calls and bytes.

Now for the initialisation you need to set up calls with a the address of a variable with value zero. There is probably a better way to do that, but this passes.

There is rather a lot of copy and paste here. Probably many improvements could be made.

","updated_at":"2018-07-13T10:01:29Z","links":{}},{"uuid":"3af61e78-08d3-11ec-81c5-853579c25f94","author":{"handle":"eraserix","avatar_url":"https://assets.exercism.org/avatars/175619/0","flair":null,"reputation":"178"},"content_markdown":"I guess you (and many others) open yourself up for a race condition if you don't lock the mutex in {Read,Write}Count(). Its perfectly possible that while you copy the values in your return statement, that only one of nbytes and nops has been updated while the other has not, so you get inconsistent results. In addition, I don't think you can assume for the general case that reads and writes to int/int64 are atomic. This is, at the very least, dependant on the hardware and memory alignment.","content_html":"

I guess you (and many others) open yourself up for a race condition if you don't lock the mutex in {Read,Write}Count(). Its perfectly possible that while you copy the values in your return statement, that only one of nbytes and nops has been updated while the other has not, so you get inconsistent results. In addition, I don't think you can assume for the general case that reads and writes to int/int64 are atomic. This is, at the very least, dependant on the hardware and memory alignment.

","updated_at":"2018-07-13T10:01:29Z","links":{}}]}}},"links":{"create":"https://exercism.org/api/v2/tracks/go/exercises/paasio/community_solutions/zeimusu/comments","enable":"https://exercism.org/api/v2/tracks/go/exercises/paasio/community_solutions/zeimusu/comments/enable","disable":"https://exercism.org/api/v2/tracks/go/exercises/paasio/community_solutions/zeimusu/comments/disable"}}

Language Tracks

#48in24 Challenge

Your Journey

Exercism Perks

Community Videos

Brief Introduction Series

Interviews & Stories

Discord

Forum

Getting started

Mentoring

Docs

Contributors

Donate

About Exercism

Our Impact

Insiders

SWAG

zeimusu's solution 294 to PaaS I/O in Go

About the Go Exercise

Other solutions to PaaS I/O in Go

zeimusu's solution
294

to PaaS I/O in Go

Other solutions to PaaS I/O
in

Go