Anne spotted a bug in numberOfLinesOfCode.
In Java, one can easily find classes with numberOfLinesOfCode < numberOfLinesOfCodewithMoreThanOnecharacter
The reason is that numberOfLinesOfCodewithMoreThanOnecharacter is defined in SourcedEntity by counting the lines (with more than 1 char) in the sourceText
numberOfLinesOfCode is the sum of numberOfLinesOfCode of the methods.
This is clearly not right for Java where a class as more code than just its methods.
But it is not even coherent for Smalltalk where a class' sourceText includes the name of the class (at least 1 additional line). So even in Smalltalk, one finds classes with numberOfLinesOfCode < numberOfLinesOfCodewithMoreThanOnecharacter (happens when none of the methods of the class have blank lines e.g. FamixEntity)
So I propose to unify both methods and compute numberOfLinesOfCode in SourcedEntity from the sourceText (and remove FAMIXClass>>numberOfLinesOfCode)
Second, for methods, the same numberOfLinesOfCode has a backward compatibilty with "exporters that use LOC as a property name" (in this case, it uses LOC instead of computing the number)
Anybody still having this case?
So I propose to remove FAMIXBehaviouralEntity>>numberOfLinesOfCode
nicolas
There are many ways to calculate the number of lines of code, so I don't expect them to be consistent. Computing from the sourceText is not necessarily any better, especially in the case of smalltalk. The sourceText of a smalltalk class is not in a form where a smalltalker would find it easy to predict the #loc, or where she would find it recognizable as a smalltalk class (missing package and superclass)
e.g. Smalltalk::Job CIV#jobAnnouncer block currentValue min max title children isRunning parent process ...
I would either test-drive the different ways of calculating #loc describing what exactly we mean, and making those different calculations available, or not touch it.
There are a.o. - count empty lines - count empty lines beyond end - count empty lines before class start - count license comment - reduce multiple empty lines to one - count method separators - wrap - word-wrap - ignore comments - ignore multi-line comments - map instance variables from string to separate lines - package name
Stephan
On 01/01/17 13:18, Nicolas Anquetil wrote:
Anne spotted a bug in numberOfLinesOfCode.
...
So I propose to remove FAMIXBehaviouralEntity>>numberOfLinesOfCode
yes there are many ways
but I don't see any reasonable argument for numberOfLinesOfCode to be inferior to numberOfLinesOfCodWithMorethanOneCharacter
nicolas
On 01/01/2017 16:05, stephan wrote:
There are many ways to calculate the number of lines of code, so I don't expect them to be consistent. Computing from the sourceText is not necessarily any better, especially in the case of smalltalk. The sourceText of a smalltalk class is not in a form where a smalltalker would find it easy to predict the #loc, or where she would find it recognizable as a smalltalk class (missing package and superclass)
e.g. Smalltalk::Job CIV#jobAnnouncer block currentValue min max title children isRunning parent process ...
I would either test-drive the different ways of calculating #loc describing what exactly we mean, and making those different calculations available, or not touch it.
There are a.o.
- count empty lines
- count empty lines beyond end
- count empty lines before class start
- count license comment
- reduce multiple empty lines to one
- count method separators
- wrap
- word-wrap
- ignore comments
- ignore multi-line comments
- map instance variables from string to separate lines
- package name
Stephan
On 01/01/17 13:18, Nicolas Anquetil wrote:
Anne spotted a bug in numberOfLinesOfCode.
...
So I propose to remove FAMIXBehaviouralEntity>>numberOfLinesOfCode
Moose-dev mailing list Moose-dev@list.inf.unibe.ch https://www.list.inf.unibe.ch/listinfo/moose-dev
-
Nicolas Anquetil -
stephan