Hi Janko. Thanks for your useful package. Apparently there is some issue with #linesOfCode. The dummy expression below return 5 instead of 4 in Pharo 1.4 and 2 because is counting the autogenerated 'DoIt' selector. (UndefinedObject evaluatorClass new compileNoPattern: '| oc | oc := OrderedCollection new: 20. oc add: 1. oc size' in: UndefinedObject context: nil notifying: nil ifFail: []) generate linesOfCode Hernán El 31/03/2013 10:17, Janko Mivšek escribió:

Hi Hernán

I once did some code statistics for our web frameworks, by counting:

- Nr of packages, categories, classes, methods - lines of code - Avg methods/class - Avg lines/method

- Nr of test methods - Avg tests/class

- % of commented methods - % of commented classes

Then three graphs were drawn:

- Methods in classes - Class inheritance depth - Lines in methods

Results can be seen here:

https://docs.google.com/spreadsheet/ccc?key=0Atnb1W9vuq9ndG4wbUpxNXpKQnFZRlF...

The code measuring all above is here:

http://www.squeaksource.com/WebCodeStats.html

Hope this helps Janko

Dne 31. 03. 2013 09:23, piše Hernán Morales Durand:

...

Hi there, Guys, what are you using to measure logical lines of code (LOC) in Pharo? Given I am measuring a script, is there any other software measure you may recommend? Cheers,

Hernán

On Apr 2, 2013, at 4:54 AM, Hernán Morales Durand <hernan.morales@gmail.com> wrote:

Hi Janko. Thanks for your useful package.

Apparently there is some issue with #linesOfCode. The dummy expression below return 5 instead of 4 in Pharo 1.4 and 2 because is counting the autogenerated 'DoIt' selector.

(UndefinedObject evaluatorClass new compileNoPattern: '| oc | oc := OrderedCollection new: 20. oc add: 1. oc size' in: UndefinedObject context: nil notifying: nil ifFail: []) generate linesOfCode

Hello, I do not think that this is wrong. The doit *does* contain a selector. if #linesOfCode would contain a special case for doits, it would be inconsistent. E.g. print this: thisContext method sourceCode --> 'DoIt ^ thisContext method sourceCode' There is a Doit. There is even a return in the that is generated… and the doit is actually pretty-printed: thisContext method linesOfCode --> 2 So in short: dots are special, there is no need to change #linesOfCode, as there is no thinkable change that would make any sense. In general, LOC as a metric is not that good. It is better than nothing, but it makes more sense to count e.g. message sends or use more advanced metrics on the AST (not every statement adds the same to complexity, e.g. conditionals versus simple statements). Marcus

Perhaps you could give Moose a try :). For a quick experiment: 1. Download https://ci.inria.fr/moose/job/moose-latest-dev-4.8/lastSuccessfulBuild/artif... 2. Execute ((MoosePharoImporterTask new addClasses: {Morph}; run) allModelClasses entityNamed: #'Smalltalk::Morph') openInMoose 3. Look in the second tab Cheers, Doru On Apr 7, 2013, at 11:01 PM, Hernán Morales Durand <hernan.morales@gmail.com> wrote:

Hi Marcus,

Thanks for the information. I couldn't answer before because of the flooding in my city[*].

Can you recommend a package for more advanced metrics? Cheers,

[*] http://www.bbc.co.uk/news/world-latin-america-22023196 http://www.bbc.co.uk/news/world-latin-america-22019841

El 02/04/2013 11:01, Marcus Denker escribió:

...

On Apr 2, 2013, at 4:54 AM, Hernán Morales Durand <hernan.morales@gmail.com> wrote:

...

Hi Janko. Thanks for your useful package.

Apparently there is some issue with #linesOfCode. The dummy expression below return 5 instead of 4 in Pharo 1.4 and 2 because is counting the autogenerated 'DoIt' selector.

(UndefinedObject evaluatorClass new compileNoPattern: '| oc | oc := OrderedCollection new: 20. oc add: 1. oc size' in: UndefinedObject context: nil notifying: nil ifFail: []) generate linesOfCode

Hello,

I do not think that this is wrong. The doit *does* contain a selector. if #linesOfCode would contain a special case for doits, it would be inconsistent.

E.g. print this:

thisContext method sourceCode

-->

'DoIt ^ thisContext method sourceCode'

There is a Doit. There is even a return in the that is generated… and the doit is actually pretty-printed:

thisContext method linesOfCode

--> 2

So in short: dots are special, there is no need to change #linesOfCode, as there is no thinkable change that would make any sense.

In general, LOC as a metric is not that good. It is better than nothing, but it makes more sense to count e.g. message sends or use more advanced metrics on the AST (not every statement adds the same to complexity, e.g. conditionals versus simple statements).

Marcus

-- www.tudorgirba.com "Speaking louder won't make the point worthier."