On Class and Method comments

Now, I only know that comments in block closures are placed behind the block code.
Do you know more examplmes where the formatter still misplaces the comment.

Maybe just comments now? I've added a formatting configuration that preserved vertical white space (and I believe also comment position, because I used to write inline comments), but I think it was lost during migration to BlueInk. But I'll have a look.

 

On Tue, May 17, 2016 at 5:32 PM, Jimmie Houchin <[hidden email]> wrote:

On 05/17/2016 01:40 AM, Peter Uhnák wrote:

As a note, maybe try to avoid inline comments unless formatter is fixed, because autoformatting _will_ misplace them. (Not to mention that imho inline comments in Smalltalk are a result of a bad design… if you need an inline comment, maybe turn that into actual code).

I disagree with the comment about inline comments being bad design. At least as a broadly applied as the above was.

Comments often say "Why" we did something. Code says "How" or "What" we are doing. I have encountered many times where if I did not have a comment for the "Why". Then at a later date when I don't remember the "Why". I encounter code that looks strange and I am tempted to refactor and consequently would introduce bugs.

This doesn't necessarily dispute my point. If you need to have inline comments then you are most likely doing in the method more then you should, otherwise you can just put the intent in the method comment.

Of course in real world we have looooong methods in both ancient and (unfortunately) new systems, and adding a comment is better then nothing.

Assuming of course, that the comment is correct… because a wrong or misleading comment can do more harm than good… that's why I _personally_ (thus the imho in the original note) prefer code over comments. Because code is what actually happens.

And finally I rarely (unless the method is a real mess) have problem understanding a single method (I can poke it, look at tests, go through the code, ...). What I usually lack is high-level overview that would tell me how the system overall works. That you cannot read from method (or class) comments.

 

Right now I am working on parsing a bunch of xml files. Third party data sources.

Is it XML or XMI? Couple of weeks back a wrote a simple utility for Synectique to help with analyzing and processing XMI files. Maybe it could help you a bit https://github.com/peteruhnak/xmi-analyzer ?

 

There is ugliness in the world. I have to clean the ugliness in order to be able to parse the files. There is no option if I want to do this. I have no control over the sources. I am not interested in engaging outside bureaucratic policies and people in order to change their procedures and policies so that I can have better sources. How often would I be willing to fight this losing battle.

The only solution is to do ugly things, for good reasons and comment them well. So that I know not just what I did, but why.

I also believe there is some conflicting policy in the code critic or whatever is telling me my method is too long.

I have frequently written methods that had a few temp vars with nice informative names. and the method is involved enough to use these vars multiple times. I get this nice information informing me my method is long. Or in Pharo4 going yellow and then red. :)

However many times if I made these vars short, like x, y, z. Then the method is a perfectly fine length. Ugh. I am then being punished for length of names. I have seen this a lot.

Well Code Critic doesn't force you to do all these things. But every time you transgress the rule you should consider whether it's worth it. (And maybe even auto-ignore the rule.)

 

In any case, there are no guidelines afaik… otherwise we would have comments everywhere already. :)

Not necessarily. People are busy. Sometimes getting working code out is more important than commenting it well.

For example. And I am not saying this is true. It is simply and example. While pressure is mounting to release Pharo 5. The few over worked people involved in that final process. Finishing up the final things, might be pragmatic and let working code go without comments. Or in some similar situation by a code contributor.

Then their are people like myself and Dimitris who might not feel confident in submitting comments for some of these areas without appropriate guidelines. We want to improve the situation not make bad examples of what not to do. :)

Just some of my opinions.

And finally there is never one rule that would fit all, and any guideline or whatever will find an edge case. But Pharo is moving fast… so I don't see why you wouldn't feel confident about commenting. The worst case scenario is that something breaks and we have a chance to learn something from the mistake.

Peter

 

Jimmie

Perhaps one thing regarding class comments: There was a push to use PIllar there if you need formatting, but I am not sure what is the current status of that.

Peter

On Tue, May 17, 2016 at 12:04 AM, Dimitris Chloupis <[hidden email][hidden email]> wrote:

I really like to start writing some class and method comments to make Pharo image more beginner friendly. Are there any guidelines when and how to comment classes and methods ? What about inline comments ?

You are absolutely right. Many without comments. Some of which may not need comments as they are self revealing in what and why. But many which do need some good comments.

As I read Peter's email. What he was saying is that my intent or why can be placed in the method comment when broken into smaller methods. Instead of it being inline in a larger method. Sometimes that is true.

What I am encountering sometimes at the moment in parsing the XML is ugly XML source. I have methods with way too many ifTrue: [] ifFalse: [ ifTrue: [] ifFalse: []] ... nested conditionals. Sometimes this is not easy to break up into smaller methods. All of the state is in the nested conditionals.

In this situation it is occurring as I parse the 16MB XML file and encounter situations in the source. The debugger pops up. I have to deal with whatever. Sometimes this needs some inline comments. Sometimes after I have the code working. I may or may not be able to reasonably refactor into something nicer and cleaner. Sometimes it is more important to move on to other more pressing issues.

Sometimes I think this why we have uncommented or poorly commented code in the image. It is working. I need to move on. Really we want some of the big movers to have this liberty. And allow some of us to follow behind, learn and comment.

Thank you for this effort to help improve our home (image). :)
I need to find time and learn to contribute also.

Jimmie

On 05/17/2016 11:59 AM, Dimitris Chloupis wrote:

ok guys then it seems all is as expected

I am a fan of inline comments when needed, I am with full agreement with Jimmie here , braking methods to smaller methods cannot reveal intend, an inline comment can but thats not big deal i can put everything on the start so everyone is happy

The problem here is that there many classes with no comments and methods that inline comments is the least of my worries.

On Tue, May 17, 2016 at 7:20 PM Peter Uhnák <[hidden email]> wrote:

Now, I only know that comments in block closures are placed behind the block code.
Do you know more examplmes where the formatter still misplaces the comment.

Maybe just comments now? I've added a formatting configuration that preserved vertical white space (and I believe also comment position, because I used to write inline comments), but I think it was lost during migration to BlueInk. But I'll have a look.

 

On Tue, May 17, 2016 at 5:32 PM, Jimmie Houchin <[hidden email]> wrote:

On 05/17/2016 01:40 AM, Peter Uhnák wrote:

As a note, maybe try to avoid inline comments unless formatter is fixed, because autoformatting _will_ misplace them. (Not to mention that imho inline comments in Smalltalk are a result of a bad design… if you need an inline comment, maybe turn that into actual code).

I disagree with the comment about inline comments being bad design. At least as a broadly applied as the above was.

Comments often say "Why" we did something. Code says "How" or "What" we are doing. I have encountered many times where if I did not have a comment for the "Why". Then at a later date when I don't remember the "Why". I encounter code that looks strange and I am tempted to refactor and consequently would introduce bugs.

This doesn't necessarily dispute my point. If you need to have inline comments then you are most likely doing in the method more then you should, otherwise you can just put the intent in the method comment.

Of course in real world we have looooong methods in both ancient and (unfortunately) new systems, and adding a comment is better then nothing.

Assuming of course, that the comment is correct… because a wrong or misleading comment can do more harm than good… that's why I _personally_ (thus the imho in the original note) prefer code over comments. Because code is what actually happens.

And finally I rarely (unless the method is a real mess) have problem understanding a single method (I can poke it, look at tests, go through the code, ...). What I usually lack is high-level overview that would tell me how the system overall works. That you cannot read from method (or class) comments.

 

Right now I am working on parsing a bunch of xml files. Third party data sources.

Is it XML or XMI? Couple of weeks back a wrote a simple utility for Synectique to help with analyzing and processing XMI files. Maybe it could help you a bit https://github.com/peteruhnak/xmi-analyzer ?

 

There is ugliness in the world. I have to clean the ugliness in order to be able to parse the files. There is no option if I want to do this. I have no control over the sources. I am not interested in engaging outside bureaucratic policies and people in order to change their procedures and policies so that I can have better sources. How often would I be willing to fight this losing battle.

The only solution is to do ugly things, for good reasons and comment them well. So that I know not just what I did, but why.

I also believe there is some conflicting policy in the code critic or whatever is telling me my method is too long.

I have frequently written methods that had a few temp vars with nice informative names. and the method is involved enough to use these vars multiple times. I get this nice information informing me my method is long. Or in Pharo4 going yellow and then red. :)

However many times if I made these vars short, like x, y, z. Then the method is a perfectly fine length. Ugh. I am then being punished for length of names. I have seen this a lot.

Well Code Critic doesn't force you to do all these things. But every time you transgress the rule you should consider whether it's worth it. (And maybe even auto-ignore the rule.)

 

In any case, there are no guidelines afaik… otherwise we would have comments everywhere already. :)

Not necessarily. People are busy. Sometimes getting working code out is more important than commenting it well.

For example. And I am not saying this is true. It is simply and example. While pressure is mounting to release Pharo 5. The few over worked people involved in that final process. Finishing up the final things, might be pragmatic and let working code go without comments. Or in some similar situation by a code contributor.

Then their are people like myself and Dimitris who might not feel confident in submitting comments for some of these areas without appropriate guidelines. We want to improve the situation not make bad examples of what not to do. :)

Just some of my opinions.

And finally there is never one rule that would fit all, and any guideline or whatever will find an edge case. But Pharo is moving fast… so I don't see why you wouldn't feel confident about commenting. The worst case scenario is that something breaks and we have a chance to learn something from the mistake.

Peter

 

Jimmie

Perhaps one thing regarding class comments: There was a push to use PIllar there if you need formatting, but I am not sure what is the current status of that.

Peter

On Tue, May 17, 2016 at 12:04 AM, Dimitris Chloupis <[hidden email]> wrote:

I really like to start writing some class and method comments to make Pharo image more beginner friendly. Are there any guidelines when and how to comment classes and methods ? What about inline comments ?

I really like to start writing some class and method comments to make Pharo image more beginner friendly.

This would be realllllllllly coool.
Did you check the class comment template because I wrote it thinking about this.

What I think would be great is to have great comments for the basic classes. Because as a newby such classes are really important.

For example I would like to enhance the symbol class comment to include the snippet sent by henrik

| s1 s2 |
s1 := 1234 asString. s2 := 1234 asString.
s1 = s2. "true" s1 == s2. "false"
s1 asSymbol = s2 asSymbol. "true" s1 asSymbol == s2 asSymbol. "true"
(s1 class allInstances select: [:s | s = s1 ]) size. "2" (s1 asSymbol class allInstances select: [:s | s = s1 asSymbol ]) size. "1"
[ #stringA = #stringB ] bench. "26,812,864 per second" [ 'StringA' = 'StringB' ] bench. "3,492,987 per second"

Are there any guidelines when and how to comment classes and methods ? What about inline comments ?