Smalltalk › Pharo › Pharo Smalltalk Developers

Strange DNU effect on Windows with Pastell

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

14 messages Options

Torsten Bergmann

Strange DNU effect on Windows with Pastell

I have a strange effect I was not able to find out so far. You can try with
any newer Pharo image/VM but here are the exact combination I tried:

1. Load latest VM from https://dl.bintray.com/pharo-project/pharo-vm/
which is as of today
(https://dl.bintray.com/pharo-project/pharo-vm/pharo-win-i386-201701091105-b0400d4.zip)

Extract in a directory.

2. Download and extract http://files.pharo.org/sources/PharoV50.sources.zip into same directory

3. Use latest image http://files.pharo.org/image/60/60342.zip in same directory

4. Start Pharo and load "Pastell"

Metacello new
smalltalkhubUser: 'Pharo' project: 'MetaRepoForPharo50';
configuration: 'Pastell';
version: #stable;
load.

5. Put some Transcript output into: XMLNodeWithElements>>#doesNotUnderstand: by changing it
to:

doesNotUnderstand: aMessage
"If I have one or more childs whose name is the same as the selector,
return them in a PastellCollection.
Otherwise, let the DNU propagate"

| v |
Transcript show: aMessage selector asString;cr.
v := self child: aMessage selector.
^ v
ifEmpty: [ super doesNotUnderstand: aMessage ]
ifNotEmpty: [ v ].

6. Now evaluate:

| xml document htmls html heads |
Transcript open;clear.
xml := '<html>
<head></head>
<head></head>
</html>'.
document := XMLDOMParser parse: xml readStream.
htmls := document html. "gives a pastell collection using the #doesNotUnderstand: trick"
html := htmls first. "first element from collection, the class is XMLElement, a subclass of XMLNodeWithElements"
html head "should call #doesNotUnderstand: again in XMLNodeWithElements but it is not"

As you see the #html message triggers the #doesNotUnderstand: while #head does not.
For strange reasons the XMLNodeWithElements>>#doesNotUnderstand: is only called once and then not anymore.

Did I miss something?

Last time Pastell worked was in Pharo 3.0. Is this a Cog issue or something I do not see?
It seems to work on Mac (according to Stephan Eggermont).

Any idea?

Thanks
T.

Torsten Bergmann

Re: Strange DNU effect on Windows with Pastell

I've found it:

Loading Pastell into latest image 60342 using:

Metacello new
smalltalkhubUser: 'Pharo' project: 'MetaRepoForPharo50';
configuration: 'Pastell';
version: #stable;
load.

the tests are green now.

And it is also not a difference between Mac and Windows

When the method looks like

testDNU
| xml document html testElements |
xml := '<html><testElement/><testElement/></html>'.
document := XMLDOMParser parse: xml readStream.
html := document html.
testElements := html testElement.
self assert: testElements size = 2.
self assert: testElements first name = #testElement.
self assert: testElements last name = #testElement

it works

When I exchange the tag "testElement" agains "head" it does not work as I reported
in my last mail:

testDNU
| xml document html testElements |
xml := '<html><head/><head/></html>'.
document := XMLDOMParser parse: xml readStream.
html := document html.
testElements := html head.
self assert: testElements size = 2.
self assert: testElements first name = #head.
self assert: testElements last name = #head

Now the assertions fail although only the tag name/message changed.

Reason is simple: there is a #head method in Object from Morphic just returning
self. For whatever reason but this makes the difference.

Lesson learned: take care with DNU tricks ;)

Sorry for the noise!

Bye
T.

Sven Van Caekenberghe-2

Re: Strange DNU effect on Windows with Pastell

> On 11 Jan 2017, at 09:33, Torsten Bergmann <[hidden email]> wrote:
>
> I've found it:
>
> Loading Pastell into latest image 60342 using:
>
>
> Metacello new
> smalltalkhubUser: 'Pharo' project: 'MetaRepoForPharo50';
> configuration: 'Pastell';
> version: #stable;
> load.
>
> the tests are green now.
>
> And it is also not a difference between Mac and Windows
>
>
> When the method looks like
>
> testDNU
> | xml document html testElements |
> xml := '<html><testElement/><testElement/></html>'.
> document := XMLDOMParser parse: xml readStream.
> html := document html.
> testElements := html testElement.
> self assert: testElements size = 2.
> self assert: testElements first name = #testElement.
> self assert: testElements last name = #testElement
>
> it works
>
>
> When I exchange the tag "testElement" agains "head" it does not work as I reported
> in my last mail:
>
> testDNU
> | xml document html testElements |
> xml := '<html><head/><head/></html>'.
> document := XMLDOMParser parse: xml readStream.
> html := document html.
> testElements := html head.
> self assert: testElements size = 2.
> self assert: testElements first name = #head.
> self assert: testElements last name = #head
>
> Now the assertions fail although only the tag name/message changed.
>
> Reason is simple: there is a #head method in Object from Morphic just returning
> self. For whatever reason but this makes the difference.
>
> Lesson learned: take care with DNU tricks ;)
>
> Sorry for the noise!

That's not noise, it is an interesting lesson !

But why the hell would there be a Object>>#head ?!

IMHO such general selectors at this level should not be used unless there is a very good reason to do so.

> Bye
> T.
>
>

Torsten Bergmann

Re: Strange DNU effect on Windows with Pastell

> But why the hell would there be a Object>>#head ?!
>
> IMHO such general selectors at this level should not be used unless there is a very good reason to do so.
>

Yes - some Morphic addition. Version says: BenjaminVanRyseghem in 2013
Only used in MorphTreeNodeMorph and to me this smells. We should nuke it.

Thx
T.

Sven Van Caekenberghe-2

Re: Strange DNU effect on Windows with Pastell

> On 11 Jan 2017, at 09:55, Torsten Bergmann <[hidden email]> wrote:
>
>> But why the hell would there be a Object>>#head ?!
>>
>> IMHO such general selectors at this level should not be used unless there is a very good reason to do so.
>>
>
> Yes - some Morphic addition. Version says: BenjaminVanRyseghem in 2013
> Only used in MorphTreeNodeMorph and to me this smells. We should nuke it.
>
> Thx
> T.

Looking in a 6 image, there is also an even uglier implementation in

Association>>#head

^ (key isKindOf: Association)
ifTrue: [ key head ]
ifFalse: [ key ]

Duh.

And I can't seen any senders...

EstebanLM

Re: Strange DNU effect on Windows with Pastell

at least we should rename it as “treeMorphHead”… just “head” is too generic.

Esteban

> On 11 Jan 2017, at 10:03, Sven Van Caekenberghe <[hidden email]> wrote:
>
>
>> On 11 Jan 2017, at 09:55, Torsten Bergmann <[hidden email]> wrote:
>>
>>> But why the hell would there be a Object>>#head ?!
>>>
>>> IMHO such general selectors at this level should not be used unless there is a very good reason to do so.
>>>
>>
>> Yes - some Morphic addition. Version says: BenjaminVanRyseghem in 2013
>> Only used in MorphTreeNodeMorph and to me this smells. We should nuke it.
>>
>> Thx
>> T.
>
> Looking in a 6 image, there is also an even uglier implementation in
>
> Association>>#head
>
> ^ (key isKindOf: Association)
> ifTrue: [ key head ]
> ifFalse: [ key ]
>
> Duh.
>
> And I can't seen any senders...
>
>

EstebanLM

Re: Strange DNU effect on Windows with Pastell

> On 11 Jan 2017, at 10:12, Esteban Lorenzano <[hidden email]> wrote:
>
> at least we should rename it as “treeMorphHead”… just “head” is too generic.

“treeNodeHead” I meant.

>
> Esteban
>
>
>> On 11 Jan 2017, at 10:03, Sven Van Caekenberghe <[hidden email]> wrote:
>>
>>
>>> On 11 Jan 2017, at 09:55, Torsten Bergmann <[hidden email]> wrote:
>>>
>>>> But why the hell would there be a Object>>#head ?!
>>>>
>>>> IMHO such general selectors at this level should not be used unless there is a very good reason to do so.
>>>>
>>>
>>> Yes - some Morphic addition. Version says: BenjaminVanRyseghem in 2013
>>> Only used in MorphTreeNodeMorph and to me this smells. We should nuke it.
>>>
>>> Thx
>>> T.
>>
>> Looking in a 6 image, there is also an even uglier implementation in
>>
>> Association>>#head
>>
>> ^ (key isKindOf: Association)
>> ifTrue: [ key head ]
>> ifFalse: [ key ]
>>
>> Duh.
>>
>> And I can't seen any senders...
>>
>>
>

Henrik-Nergaard

Re: Strange DNU effect on Windows with Pastell

And Object/Association >> #tail

Best regards,
Henrik

-----Opprinnelig melding-----
Fra: Pharo-dev [mailto:[hidden email]] På vegne av Esteban Lorenzano
Sendt: 11 January 2017 10:13
Til: Pharo Development List <[hidden email]>
Emne: Re: [Pharo-dev] Strange DNU effect on Windows with Pastell

> On 11 Jan 2017, at 10:12, Esteban Lorenzano <[hidden email]> wrote:
>
> at least we should rename it as “treeMorphHead”… just “head” is too generic.

“treeNodeHead” I meant.

>
> Esteban
>
>
>> On 11 Jan 2017, at 10:03, Sven Van Caekenberghe <[hidden email]> wrote:
>>
>>
>>> On 11 Jan 2017, at 09:55, Torsten Bergmann <[hidden email]> wrote:
>>>
>>>> But why the hell would there be a Object>>#head ?!
>>>>
>>>> IMHO such general selectors at this level should not be used unless there is a very good reason to do so.
>>>>
>>>
>>> Yes - some Morphic addition. Version says: BenjaminVanRyseghem in
>>> 2013 Only used in MorphTreeNodeMorph and to me this smells. We should nuke it.
>>>
>>> Thx
>>> T.
>>
>> Looking in a 6 image, there is also an even uglier implementation in
>>
>> Association>>#head
>>
>> ^ (key isKindOf: Association)
>> ifTrue: [ key head ]
>> ifFalse: [ key ]
>>
>> Duh.
>>
>> And I can't seen any senders...
>>
>>
>

EstebanLM

Re: Strange DNU effect on Windows with Pastell

in general, rule of thumb when extending classes should be “use the most specific name possible”, for instance: #magritteDescription instead of #description, etc.

but everybody makes mistakes… in voyage I added methods #save, #remove instead a #voyageSave, #voyageRemove… one could argue that since that is *not* part of the image is ok, but no: is a framework so possibility of collisions are always there (of course, when doing your own project methods is another things… and I’m sure I can find other exceptions… but as I said … “in general”)

Esteban

> On 11 Jan 2017, at 11:46, Henrik Nergaard <[hidden email]> wrote:
>
> And Object/Association >> #tail
>
> Best regards,
> Henrik
>
> -----Opprinnelig melding-----
> Fra: Pharo-dev [mailto:[hidden email]] På vegne av Esteban Lorenzano
> Sendt: 11 January 2017 10:13
> Til: Pharo Development List <[hidden email]>
> Emne: Re: [Pharo-dev] Strange DNU effect on Windows with Pastell
>
>
>> On 11 Jan 2017, at 10:12, Esteban Lorenzano <[hidden email]> wrote:
>>
>> at least we should rename it as “treeMorphHead”… just “head” is too generic.
>
> “treeNodeHead” I meant.
>
>>
>> Esteban
>>
>>
>>> On 11 Jan 2017, at 10:03, Sven Van Caekenberghe <[hidden email]> wrote:
>>>
>>>
>>>> On 11 Jan 2017, at 09:55, Torsten Bergmann <[hidden email]> wrote:
>>>>
>>>>> But why the hell would there be a Object>>#head ?!
>>>>>
>>>>> IMHO such general selectors at this level should not be used unless there is a very good reason to do so.
>>>>>
>>>>
>>>> Yes - some Morphic addition. Version says: BenjaminVanRyseghem in
>>>> 2013 Only used in MorphTreeNodeMorph and to me this smells. We should nuke it.
>>>>
>>>> Thx
>>>> T.
>>>
>>> Looking in a 6 image, there is also an even uglier implementation in
>>>
>>> Association>>#head
>>>
>>> ^ (key isKindOf: Association)
>>> ifTrue: [ key head ]
>>> ifFalse: [ key ]
>>>
>>> Duh.
>>>
>>> And I can't seen any senders...
>>>
>>>
>>
>
>

Sven Van Caekenberghe-2

Re: Strange DNU effect on Windows with Pastell

In reply to this post by Henrik-Nergaard

All good and well, but maybe its me, but I can't see where it is being used ...

Henrik-Nergaard

Re: Strange DNU effect on Windows with Pastell

Both #head and #tail is used by MorphTreeNodeMorph (#expandPath: and #matchPath:) to select items deeper inside the tree structure and then expanding to this.
-------------------------------------------------------------------------
| model morph |
model := ClassTreeExample new.
model openOn: Morph.
model selectItems: { Morph -> BorderedMorph -> AlignmentMorph }.
-------------------------------------------------------------------------

Changing one should not be done without the other.

Best regards,
Henrik

-----Opprinnelig melding-----
Fra: Pharo-dev [mailto:[hidden email]] På vegne av Sven Van Caekenberghe
Sendt: 11 January 2017 11:56
Til: Pharo Development List <[hidden email]>
Emne: Re: [Pharo-dev] Strange DNU effect on Windows with Pastell

All good and well, but maybe its me, but I can't see where it is being used ...

Sven Van Caekenberghe-2

Re: Strange DNU effect on Windows with Pastell

OK, thanks, you are right; I was confused by the new filter option in senders, somehow I assumed the list was not longer (and I ignored the scroll bar) - weird. I see them now.

Yes, both should be changed together.

> On 11 Jan 2017, at 12:33, Henrik Nergaard <[hidden email]> wrote:
>
> Both #head and #tail is used by MorphTreeNodeMorph (#expandPath: and #matchPath:) to select items deeper inside the tree structure and then expanding to this.
> -------------------------------------------------------------------------
> | model morph |
> model := ClassTreeExample new.
> model openOn: Morph.
> model selectItems: { Morph -> BorderedMorph -> AlignmentMorph }.
> -------------------------------------------------------------------------
>
> Changing one should not be done without the other.
>
> Best regards,
> Henrik
>
> -----Opprinnelig melding-----
> Fra: Pharo-dev [mailto:[hidden email]] På vegne av Sven Van Caekenberghe
> Sendt: 11 January 2017 11:56
> Til: Pharo Development List <[hidden email]>
> Emne: Re: [Pharo-dev] Strange DNU effect on Windows with Pastell
>
> All good and well, but maybe its me, but I can't see where it is being used ...
>
>> On 11 Jan 2017, at 11:46, Henrik Nergaard <[hidden email]> wrote:
>>
>> And Object/Association >> #tail
>>
>> Best regards,
>> Henrik
>>
>> -----Opprinnelig melding-----
>> Fra: Pharo-dev [mailto:[hidden email]] På vegne av Esteban Lorenzano
>> Sendt: 11 January 2017 10:13
>> Til: Pharo Development List <[hidden email]>
>> Emne: Re: [Pharo-dev] Strange DNU effect on Windows with Pastell
>>
>>
>>> On 11 Jan 2017, at 10:12, Esteban Lorenzano <[hidden email]> wrote:
>>>
>>> at least we should rename it as “treeMorphHead”… just “head” is too generic.
>>
>> “treeNodeHead” I meant.
>>
>>>
>>> Esteban
>>>
>>>
>>>> On 11 Jan 2017, at 10:03, Sven Van Caekenberghe <[hidden email]> wrote:
>>>>
>>>>
>>>>> On 11 Jan 2017, at 09:55, Torsten Bergmann <[hidden email]> wrote:
>>>>>
>>>>>> But why the hell would there be a Object>>#head ?!
>>>>>>
>>>>>> IMHO such general selectors at this level should not be used unless there is a very good reason to do so.
>>>>>>
>>>>>
>>>>> Yes - some Morphic addition. Version says: BenjaminVanRyseghem in
>>>>> 2013 Only used in MorphTreeNodeMorph and to me this smells. We should nuke it.
>>>>>
>>>>> Thx
>>>>> T.
>>>>
>>>> Looking in a 6 image, there is also an even uglier implementation in
>>>>
>>>> Association>>#head
>>>>
>>>> ^ (key isKindOf: Association)
>>>> ifTrue: [ key head ]
>>>> ifFalse: [ key ]
>>>>
>>>> Duh.
>>>>
>>>> And I can't seen any senders...
>>>>
>>>>
>>>
>>
>>
>
>

stepharong

Re: Strange DNU effect on Windows with Pastell

In reply to this post by Torsten Bergmann

can you open a bug entry.
I'm dreaming about a Object with far less funky methods.

On Wed, 11 Jan 2017 09:55:06 +0100, Torsten Bergmann <[hidden email]>
wrote:

>> But why the hell would there be a Object>>#head ?!
>>
>> IMHO such general selectors at this level should not be used unless
>> there is a very good reason to do so.
>>
>
> Yes - some Morphic addition. Version says: BenjaminVanRyseghem in 2013
> Only used in MorphTreeNodeMorph and to me this smells. We should nuke it.
>
> Thx
> T.
>

--
Using Opera's mail client: http://www.opera.com/mail/

stepharong

Re: Strange DNU effect on Windows with Pastell

In reply to this post by Sven Van Caekenberghe-2

On Wed, 11 Jan 2017 12:48:43 +0100, Sven Van Caekenberghe <[hidden email]>
wrote:

> OK, thanks, you are right; I was confused by the new filter option in
> senders, somehow I assumed the list was not longer (and I ignored the
> scroll bar) - weird./mail/

yes it happens to me a lot
and it is annoying.