[Cuis] Strategy for porting packages (was Re: CSV Parser in Cuis?)

Sat May 9 13:50:54 CDT 2015

On 5/9/15, H. Hirzel <hannes.hirzel at gmail.com> wrote:
> On 5/9/15, Juan Vuletich <juan at jvuletich.org> wrote:
>> Well, this issue es pretty likely to happen when porting packages. There
>> are (that I see) 3 alternatives:
>>
>> 1) Add missing methods/classes to Cuis
>> 2) Add missing methods/classes to SqueakCompatibility.pck.st (also
>> suitable for compatibility with Pharo)
>> 3) Modify the package to use existing facilities
>>
>> There are no rules written in stone, and I think we need to decide in
>> each case.
>>
>> For example, when the missing stuff is the kind of extra complexity we
>> don't want in Cuis, or we don't want it for any reason, this precludes
>> 1). But if the missing stuff fits nicely into Cuis, then the appropriate
>> choice is 1)
>>
>> If who's doing the port is also adopting the code, because (s)he wants
>> to do so, or is very interested in it, or is going to further develop
>> it, then 3) is a reasonable choice. But if the original version of the
>> package is under heavy development, and we are not really adopting it,
>> then 3) is not a good option. For instance, these days I ported VMMaker
>> to Cuis (will publish it this weekend). VMMaker is very complex, and is
>> in active development by dedicated and knowledgeable people. So, 3) is
>> out of the question. But it uses many things that don't fit Cuis, some
>> for purely aesthetic reasons, others because full compatibility would
>> turn Cuis again into Squeak. So I added quite several things to
>> SqueakCompatibility.pck.st, option 2). This made it possible to run
>> VMMaker in Cuis (without the GUI), without modifying a single line of
>> code and without Undeclared. Rebasing on future releases of VMMaker will
>> not be a problem at all.
>>
>> And this needs to be thought for each class or method missing. When
>> porting a package, maybe we choose 1) for some methods and 2) for
>> others. Or even yet 3) for others, if we are adopting the code.
>>
>> So, in this case, Hannes, please suggest, and lets discuss from there.
>
> Thank you for outlining the porting strategy alternatives. To answer
> your question about porting the Neo-CSV package from Pharo I'll have
> to do more investigations for this particular case.
>
> For example the testing code uses besides join: (for Strings) a Pharo
> 4.0 method
>
> Object>assert: aBlock description: aStringOrBlock
> 	"Throw an assertion error if aBlock does not evaluates to true."
>
> --Hannes

Here is the analysis

https://github.com/hhzl/Cuis-NeoCSV/blob/master/README.md

It turns out that the solution is a combination of all three approaches.

1) Add missing methods/classes to Cuis
2) Add missing methods/classes to SqueakCompatibility.pck.st (also
suitable for compatibility with Pharo)
3) Modify the package to use existing facilities

There are three github issues, a discussion there at the level of the
issue or here is welcome.

--Hannes

>> Cheers,
>> Juan Vuletich
>>
>> On 09/05/2015 11:58 a.m., H. Hirzel wrote:
>>> Or put the question another way: Would it be fine to have a #join
>>> function is Cuis.
>>> Alternatively I can construct a test helper function which constructs
>>> the test data.
>>>
>>>
>>>
>>>
>>>
>>> On 5/9/15, H. Hirzel<hannes.hirzel at gmail.com>  wrote:
>>>> Here is the result of a first step of porting
>>>>
>>>> https://github.com/hhzl/Cuis-NeoCSV
>>>>
>>>> 48 out of 50 tests have an error. This is mainly because of the way
>>>> the test data is constructed.
>>>>
>>>> input := (String crlf join: #( '"x","y","z"' '100,200,300'
>>>> '100,200,300' '100,200,300' '')).
>>>>
>>>>
>>>> #crlf of Pharo 4.0 may be replaced with #crlfString of Cuis, see below
>>>>
>>>> But Cuis does not have #SequencableCollection>>join
>>>>
>>>>
>>>> What are your suggestions how to proceed? It is about construction of
>>>> test
>>>> data.
>>>>
>>>>
>>>> --Hannes
>>>>
>>>>
>>>>
>>>> Pharo
>>>> crlf
>>>> 	"Answer a string containing a carriage return and a linefeed."
>>>>
>>>> 	^ self with: Character cr with: Character lf
>>>>
>>>>
>>>> Cuis
>>>>
>>>> crlfString
>>>> 	"Answer a string containing a carriage return and a linefeed."
>>>>
>>>> 	^ self with: Character cr with: Character lf
>>>>
>>>>
>>>>
>>>> Pharo class #SequencableCollection
>>>>
>>>> join: aCollection
>>>> 	"NB: this implementation only works for Array, since WriteStreams
>>>> only work for Arrays and Strings. (!)
>>>> 	Overridden in OrderedCollection and SortedCollection."
>>>> 	^ self class
>>>> 		streamContents: [:stream | aCollection
>>>> 				do: [:each | each joinTo: stream]
>>>> 				separatedBy: [stream nextPutAll: self]]
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> On 5/9/15, H. Hirzel<hannes.hirzel at gmail.com>  wrote:
>>>>> I understand that I need to click 'Install' in the file browser to get
>>>>> a separate ChangeSet object.
>>>>>
>>>>> https://github.com/Cuis-Smalltalk/Cuis-Smalltalk-Dev/blob/master/Documentation/CodeManagementInCuis.md#loading-changeset-files-into-cuis
>>>>>
>>>>>
>>>>> And it that works well and after some fixes. I do in the Package tool
>>>>>      'Create Package'
>>>>>
>>>>> https://github.com/Cuis-Smalltalk/Cuis-Smalltalk-Dev/blob/master/Documentation/CodeManagementInCuis.md#packages
>>>>>
>>>>> --Hannes
>>>>>
>>>>> On 5/9/15, H. Hirzel<hannes.hirzel at gmail.com>  wrote:
>>>>>> And it comes with documentation
>>>>>>
>>>>>> https://github.com/svenvc/docs/blob/master/neo/neo-csv-paper.md
>>>>>>
>>>>>> "CSV is a lightweight text-based de facto standard for human-readable
>>>>>> tabular data interchange."
>>>>>>
>>>>>> On 5/9/15, H. Hirzel<hannes.hirzel at gmail.com>  wrote:
>>>>>>> Yes, I assume so as well. Thank you for the note about line
>>>>>>> terminators.
>>>>>>>
>>>>>>> There is also the Pharo Neo-CSV parser which comes with many tests.
>>>>>>>
>>>>>>> https://github.com/svenvc/NeoCVS
>>>>>>>
>>>>>>> It is probably easier to take a Pharo 4.0 image, load NeoCSV in one
>>>>>>> click and file out the class categories
>>>>>>>
>>>>>>> Neo-CSV-Core
>>>>>>> Neo-CSV-Tests
>>>>>>>
>>>>>>> --Hannes
>>>>>>>
>>>>>>> On 5/9/15, Juan Vuletich<juan at jvuletich.org>  wrote:
>>>>>>>> Porting any of them shouldn't be hard. You need, as usual, to take
>>>>>>>> into
>>>>>>>> account that Squeak and Pharo uses #cr as line terminator, while
>>>>>>>> Cuis
>>>>>>>> uses #lf.
>>>>>>>>
>>>>>>>> In any case, check ExtendedClipboard.pck.st, and maybe extend it.
>>>>>>>>
>>>>>>>> Cheers,
>>>>>>>> Juan Vuletich
>>>>>>>>
>>>>>>>> On 09/05/2015 06:32 a.m., H. Hirzel wrote:
>>>>>>>>> P.S: I'd like the input to come from the clipboard. This is
>>>>>>>>>
>>>>>>>>> 1. Select a range of cells in a spreadsheet (MSExcel / Libreoffice
>>>>>>>>> calc)
>>>>>>>>> 2. Copy to clipboard
>>>>>>>>> 3. Bring the Cuis window to the front
>>>>>>>>> 4. Press a 'paste data from Excel' button
>>>>>>>>>
>>>>>>>>> In Cuis some processing will be done and the result will be put
>>>>>>>>> back
>>>>>>>>> into the clipboard
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On 5/9/15, H. Hirzel<hannes.hirzel at gmail.com>   wrote:
>>>>>>>>>> Dear all
>>>>>>>>>>
>>>>>>>>>> has anyone written or ported a CSV parser to Cuis?
>>>>>>>>>>
>>>>>>>>>> For example this one
>>>>>>>>>>
>>>>>>>>>> http://www.squeaksource.com/@BiTWvEAxOgxLE9wJ/T0wdhSm3
>>>>>>>>>>
>>>>>>>>>> Description: A simple parser for comma-separated-value files.
>>>>>>>>>> Creator:	Avi Bryant
>>>>>>>>>>
>>>>>>>>>> Regards
>>>>>>>>>>
>>>>>>>>>> Hannes Hirzel
>>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> Cuis mailing list
>>>>>>>>> Cuis at jvuletich.org
>>>>>>>>> http://jvuletich.org/mailman/listinfo/cuis_jvuletich.org
>>>>>>>>>
>>>>>>>>
>>
>>
>
> _______________________________________________
> Cuis mailing list
> Cuis at jvuletich.org
> http://jvuletich.org/mailman/listinfo/cuis_jvuletich.org
>