Fwd: Patches for Pharo packages you are maintaining

>
> On 2013-11-23, at 09:00, Stéphane Ducasse <[hidden email]> wrote:
>
>> No time so fr to handle it.
>>
>> Stef
>>
>> Begin forwarded message:
>>
>>> From: "monty" <[hidden email]>
>>> Subject: Patches for Pharo packages you are maintaining
>>> Date: November 21, 2013 3:14:25 AM GMT+01:00
>>> To: [hidden email]
>>>
>>> Prof. Stef, I have produced updates to the XML-Parser, XML-Parser-Testing, and XML-Writer-Core packages you are maintaining.
>>>
>>> These updates add validation against internal and external DTDs, proper replacement of general and parameter entites, customizable resolution of external parsed entities using Zinc and FileSystem, awareness of notations and unparsed entities, preservation of the internal DTD subset by the DOM parser (so printing a parsed doc with a DTD will produce approximately what was input), line numbers reporting in error messages, and better well-formed and validity constraints.
>>>
>>> I had to largely rewrite the tokenizer to make everything work, but I followed the spec closely, and it is about the same speed as long as there is no DTD to validate against.
>>>
>>> One problem is that while the tests I added and the existing tests all pass, for some reason helper messages in some test classes starting with "should" (in the style of should:raise:) are being interpreted as tests and run by TestRuner, even though they don't begin with "test" and take arguments! This is possibly a bug in TestRunner.
>>>
>>> Another problem is that there are so many deprecated methods cluttering up classes, some of which have been deprecated for years! It is confusing and hard to see which methods to use just by browsing the protocols. I would really suggest using this code to get rid of the XML-Parser methods that have been deprecated for at least a year:
>>>
>>> expiry := 1 year.
>>> (SystemNavigation default allClassesInPackageNamed: 'XML-Parser')
>>> do: [:class |
>>> class selectors do: [:selector | | compiledMethod timeStamp |
>>> compiledMethod := class compiledMethodAt: selector.
>>> timeStamp := compiledMethod timeStamp copyAfter: Character space.
>>> (compiledMethod isDeprecated
>>> and: [(DateAndTime now - (DateAndTime fromString: timeStamp)) > expiry])
>>> ifTrue: [class removeSelector: selector]]].
>>>
>>> I ran and checked it myself and it doesn't break anything. Running the above with XML-Writer-Core might not be a bad idea either.
>>>
>>> I also updated BitmapCharacterSet to use less memory.
>> <XML-Writer-Core-monty.5.mcz>
>> <XML-Tests-Parser-monty.27.mcz>
>> <XML-Parser-monty.173.mcz>
>> <Collections-BitmapCharacterSet-Anonymous.6.mcz>
>> <Collections-BitmapCharacterSet-Anonymous.6.mcz>
>>
>

> On 2013-11-23, at 09:00, Stéphane Ducasse <[hidden email]> wrote:
>
>> No time so fr to handle it.
>>
>> Stef
>>
>> Begin forwarded message:
>>
>>> From: "monty" <[hidden email]>
>>> Subject: Patches for Pharo packages you are maintaining
>>> Date: November 21, 2013 3:14:25 AM GMT+01:00
>>> To: [hidden email]
>>>
>>> Prof. Stef, I have produced updates to the XML-Parser, XML-Parser-Testing, and XML-Writer-Core packages you are maintaining.
>>>
>>> These updates add validation against internal and external DTDs, proper replacement of general and parameter entites, customizable resolution of external parsed entities using Zinc and FileSystem, awareness of notations and unparsed entities, preservation of the internal DTD subset by the DOM parser (so printing a parsed doc with a DTD will produce approximately what was input), line numbers reporting in error messages, and better well-formed and validity constraints.
>>>
>>> I had to largely rewrite the tokenizer to make everything work, but I followed the spec closely, and it is about the same speed as long as there is no DTD to validate against.
>>>
>>> One problem is that while the tests I added and the existing tests all pass, for some reason helper messages in some test classes starting with "should" (in the style of should:raise:) are being interpreted as tests and run by TestRuner, even though they don't begin with "test" and take arguments! This is possibly a bug in TestRunner.
>>>
>>> Another problem is that there are so many deprecated methods cluttering up classes, some of which have been deprecated for years! It is confusing and hard to see which methods to use just by browsing the protocols. I would really suggest using this code to get rid of the XML-Parser methods that have been deprecated for at least a year:
>>>
>>> expiry := 1 year.
>>> (SystemNavigation default allClassesInPackageNamed: 'XML-Parser')
>>> do: [:class |
>>> class selectors do: [:selector | | compiledMethod timeStamp |
>>> compiledMethod := class compiledMethodAt: selector.
>>> timeStamp := compiledMethod timeStamp copyAfter: Character space.
>>> (compiledMethod isDeprecated
>>> and: [(DateAndTime now - (DateAndTime fromString: timeStamp)) > expiry])
>>> ifTrue: [class removeSelector: selector]]].
>>>
>>> I ran and checked it myself and it doesn't break anything. Running the above with XML-Writer-Core might not be a bad idea either.
>>>
>>> I also updated BitmapCharacterSet to use less memory.
>> <XML-Writer-Core-monty.5.mcz>
>> <XML-Tests-Parser-monty.27.mcz>
>> <XML-Parser-monty.173.mcz>
>> <Collections-BitmapCharacterSet-Anonymous.6.mcz>
>> <Collections-BitmapCharacterSet-Anonymous.6.mcz>
>>
>

> Le 23/11/2013 19:19, Hilaire Fernandes a écrit :
>> Le 23/11/2013 19:05, Sean P. DeNigris a écrit :
>>> But seriously if this is not clear, how do we improve it?
>>>
>>
>> Make it work. It never worked for me, I keep trying then give up,
>> thinking about a temporary failure, then doing other business.
>> I don't understand what the monkey really want but it is not nice with
>> me. So I guess I miss something that should be documented.
>>
>> Hilaire
>>
>>
>
> So registering on Pharo tracker *never* worked for me, since I try,
> before the summer.
>

>
> On 23 Nov 2013, at 21:25, Camillo Bruni <[hidden email]> wrote:
>
>>
>> On 2013-11-23, at 11:02, Hilaire Fernandes <[hidden email]> wrote:
>>
>>> Le 23/11/2013 09:51, Camillo Bruni a écrit :
>>>> opened https://pharo.fogbugz.com/f/cases/12229/XML-Tools-improvement otherwise these changes are as good as lost.
>>>> sending mcz via mail is from the last century.
>>>
>>> Hello
>>>
>>> In the welcome page of the bug tracker, add a page explaining how to
>>> create by yourself an account.
>>
>> It is all there: https://pharo.fogbugz.com/default.asp?W41
>>
>> I was even so lazy that I made a video ;) since there are so many nasty steps
>> involved that I don't think text will do it.
>
> and it should be noted that the scheme we use is not the way it is because we think it’s perfect.
>
> This is really strange that people always pre-suppose that whatever is done is done with indefinite
> resources and therefore by definition exactly what was wanted. To the contrary: we just make it up as
> we go along, with very limited resources, sometimes doing wrong decisions.. and we will slowly improve.
>
> And no, just stopping until the process is perfect is *not* they way to go. This will just lead to nothing happening,
> not to perfection.

>>>
>>
>> So registering on Pharo tracker *never* worked for me, since I try,
>> before the summer.
>>
>
> In the end I think is was a mistake to use fogbugz. For sure it’s nice for companies, but
> for open sourse projects it just does not work with a closed bug tracker.
>
> Now the question is what is the effort to move to yet another one?