#valueWithPossibleArgs:, #valueWithEnoughArguments:, and #cull:
Hi: I am trying to understand the different between and perhaps origin of BlockClosure>>#valueWithPossibleArgs: and BlockClosure>>#valueWithEnoughArguments: I am trying to decide which of the two I need for SOMns. The first one has seen more recent changes, when looking at the Pharo 6.1 download: valueWithPossibleArgs: anArray â> 2/12/2017 StephaneDucasse valueWithEnoughArguments: anArray â> 3/11/2001 nk While they have rather different implementations, they seem to behave identically, as far as I could tell using the following example: blocks := { [ { } ]. [:a | { a } ]. [:a :b | { a. b } ]. [:a :b :c | { a. b. c } ] }. blocks collect: [:b | b valueWithPossibleArgs: {1}]. blocks collect: [:b | b valueWithPossibleArgs: {1. 2. 3}]. blocks collect: [:b | b valueWithEnoughArguments: {1}]. blocks collect: [:b | b valueWithEnoughArguments: {1. 2. 3}]. I was also wondering how they relate to #cull:* One of the major differences seems to be that valueWithP* and valueWithE* are both injecting nil for absent arguments, while normal #value* and #cull* methods signal an error. Is there a specific use case why one wouldnât want to be strict here as well, but instead inject nils? Any comments or pointer appreciated. Thanks Stefan -- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
Hi, It seems the two methods have exactly the same behavior indeed. valueWithPossibleArgs: anArray valueWithEnoughArguments: anArray One was edited recently but I think it's only to change the comment, they're both very old. My guess is that there are two for compatibility purpose (One is the selector that is considered as the most relevant that should be used, the other one is the one also present in other Smalltalks so we have it for cross-Smalltalk librairies or something like that), but only one is really needed. If you need only the concept for SOM-NS you can just implement one, if you want to be compatible with different Smalltalk lib implement both. All use-cases of these methods I have found do not inject nils, they expect the block to have a number of arguments of the block less or equal to the number of parameters in the argument array. I would say they're used as #cullWithArguments: but for some reason other selector names were preferred. Now, as you mentioned, these two methods are more than just cullWithArguments: since they inject nils if there are not enough parameters. To me it looks incorrect to do so because then while debugging your code you will get issues due to those injected nils and it will be tedious for the application programmer to track the problem down to these two methods. There a few use-cases for nil injection though. Typically when changing existing frameworks in multiple repositories, it may be that during the update process the change to the caller is installed before the change of the callee, and if the code is actually used (code in UI for instance), injecting nils might avoid system break-down. Another use-case is for compatibility with frameworks using the nil injection, but I can't find a framework doing that right now. Honestly, I would not implement the nil injection, but maybe some one else has a different point of view. On Fri, Feb 16, 2018 at 6:25 PM, Stefan Marr <smalltalk@stefan-marr.de> wrote:
Hi:
I am trying to understand the different between and perhaps origin of BlockClosure>>#valueWithPossibleArgs: and BlockClosure>># valueWithEnoughArguments:
I am trying to decide which of the two I need for SOMns.
The first one has seen more recent changes, when looking at the Pharo 6.1 download:
valueWithPossibleArgs: anArray â> 2/12/2017 StephaneDucasse valueWithEnoughArguments: anArray â> 3/11/2001 nk
While they have rather different implementations, they seem to behave identically, as far as I could tell using the following example:
blocks := { [ { } ]. [:a | { a } ]. [:a :b | { a. b } ]. [:a :b :c | { a. b. c } ] }.
blocks collect: [:b | b valueWithPossibleArgs: {1}]. blocks collect: [:b | b valueWithPossibleArgs: {1. 2. 3}]. blocks collect: [:b | b valueWithEnoughArguments: {1}]. blocks collect: [:b | b valueWithEnoughArguments: {1. 2. 3}].
I was also wondering how they relate to #cull:*
One of the major differences seems to be that valueWithP* and valueWithE* are both injecting nil for absent arguments, while normal #value* and #cull* methods signal an error. Is there a specific use case why one wouldnât want to be strict here as well, but instead inject nils?
Any comments or pointer appreciated.
Thanks Stefan
-- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
Hi Clement: Thanks for looking into this!
On 19 Feb 2018, at 11:43, Clément Bera <bera.clement@gmail.com> wrote:
they're both very old. My guess is that there are two for compatibility purpose
Ok, good to know. Thanks.
All use-cases of these methods I have found do not inject nils, they expect the block to have a number of arguments of the block less or equal to the number of parameters in the argument array. I would say theyâre used as #cullWithArguments: but for some reason other selector names were preferred. [...] There a few use-cases for nil injection though. Typically when changing existing frameworks in multiple repositories, it may be that during the update process the change to the caller is installed before the change of the callee, and if the code is actually used (code in UI for instance), injecting nils might avoid system break-down. Another use-case is for compatibility with frameworks using the nil injection, but I canât find a framework doing that right now.
Ok, great. In the end, I only added #cullArguments: instead of the #valueWith* variants [1]. The naming seems to communicate more clearly the expected semantics, and itâs different from the previous ones. So, in case it turns out someone needs those with nil injecting semantics, they could be added later. Thanks Stefan [1] https://github.com/smarr/SOMns/pull/215/files#diff-8dd7efb0aab92cd4b39e0bf84... -- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
Clement can you open a bug entry so that we clean this situation? Stef On Mon, Feb 19, 2018 at 12:43 PM, Clément Bera <bera.clement@gmail.com> wrote:
Hi,
It seems the two methods have exactly the same behavior indeed. valueWithPossibleArgs: anArray valueWithEnoughArguments: anArray
One was edited recently but I think it's only to change the comment, they're both very old. My guess is that there are two for compatibility purpose (One is the selector that is considered as the most relevant that should be used, the other one is the one also present in other Smalltalks so we have it for cross-Smalltalk librairies or something like that), but only one is really needed. If you need only the concept for SOM-NS you can just implement one, if you want to be compatible with different Smalltalk lib implement both.
All use-cases of these methods I have found do not inject nils, they expect the block to have a number of arguments of the block less or equal to the number of parameters in the argument array. I would say they're used as #cullWithArguments: but for some reason other selector names were preferred.
Now, as you mentioned, these two methods are more than just cullWithArguments: since they inject nils if there are not enough parameters. To me it looks incorrect to do so because then while debugging your code you will get issues due to those injected nils and it will be tedious for the application programmer to track the problem down to these two methods.
There a few use-cases for nil injection though. Typically when changing existing frameworks in multiple repositories, it may be that during the update process the change to the caller is installed before the change of the callee, and if the code is actually used (code in UI for instance), injecting nils might avoid system break-down. Another use-case is for compatibility with frameworks using the nil injection, but I can't find a framework doing that right now.
Honestly, I would not implement the nil injection, but maybe some one else has a different point of view.
On Fri, Feb 16, 2018 at 6:25 PM, Stefan Marr <smalltalk@stefan-marr.de> wrote:
Hi:
I am trying to understand the different between and perhaps origin of BlockClosure>>#valueWithPossibleArgs: and BlockClosure>>#valueWithEnoughArguments:
I am trying to decide which of the two I need for SOMns.
The first one has seen more recent changes, when looking at the Pharo 6.1 download:
valueWithPossibleArgs: anArray â> 2/12/2017 StephaneDucasse valueWithEnoughArguments: anArray â> 3/11/2001 nk
While they have rather different implementations, they seem to behave identically, as far as I could tell using the following example:
blocks := { [ { } ]. [:a | { a } ]. [:a :b | { a. b } ]. [:a :b :c | { a. b. c } ] }.
blocks collect: [:b | b valueWithPossibleArgs: {1}]. blocks collect: [:b | b valueWithPossibleArgs: {1. 2. 3}]. blocks collect: [:b | b valueWithEnoughArguments: {1}]. blocks collect: [:b | b valueWithEnoughArguments: {1. 2. 3}].
I was also wondering how they relate to #cull:*
One of the major differences seems to be that valueWithP* and valueWithE* are both injecting nil for absent arguments, while normal #value* and #cull* methods signal an error. Is there a specific use case why one wouldnât want to be strict here as well, but instead inject nils?
Any comments or pointer appreciated.
Thanks Stefan
-- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
On Thu, Feb 22, 2018 at 7:52 AM, Stephane Ducasse <stepharo.self@gmail.com> wrote:
Clement
can you open a bug entry so that we clean this situation?
What is the bug? You want to merge the two selectors and break compatibility with frameworks using the removed one?
Stef
On Mon, Feb 19, 2018 at 12:43 PM, Clément Bera <bera.clement@gmail.com> wrote:
Hi,
It seems the two methods have exactly the same behavior indeed. valueWithPossibleArgs: anArray valueWithEnoughArguments: anArray
One was edited recently but I think it's only to change the comment, they're both very old. My guess is that there are two for compatibility purpose (One is the selector that is considered as the most relevant that should be used, the other one is the one also present in other Smalltalks so we have it for cross-Smalltalk librairies or something like that), but only one is really needed. If you need only the concept for SOM-NS you can just implement one, if you want to be compatible with different Smalltalk lib implement both.
All use-cases of these methods I have found do not inject nils, they expect the block to have a number of arguments of the block less or equal to the number of parameters in the argument array. I would say they're used as #cullWithArguments: but for some reason other selector names were preferred.
Now, as you mentioned, these two methods are more than just cullWithArguments: since they inject nils if there are not enough parameters. To me it looks incorrect to do so because then while debugging your code you will get issues due to those injected nils and it will be tedious for the application programmer to track the problem down to these two methods.
There a few use-cases for nil injection though. Typically when changing existing frameworks in multiple repositories, it may be that during the update process the change to the caller is installed before the change of the callee, and if the code is actually used (code in UI for instance), injecting nils might avoid system break-down. Another use-case is for compatibility with frameworks using the nil injection, but I can't find a framework doing that right now.
Honestly, I would not implement the nil injection, but maybe some one else has a different point of view.
On Fri, Feb 16, 2018 at 6:25 PM, Stefan Marr <smalltalk@stefan-marr.de> wrote:
Hi:
I am trying to understand the different between and perhaps origin of BlockClosure>>#valueWithPossibleArgs: and BlockClosure>>#valueWithEnoughArguments:
I am trying to decide which of the two I need for SOMns.
The first one has seen more recent changes, when looking at the Pharo
6.1
download:
valueWithPossibleArgs: anArray â> 2/12/2017 StephaneDucasse valueWithEnoughArguments: anArray â> 3/11/2001 nk
While they have rather different implementations, they seem to behave identically, as far as I could tell using the following example:
blocks := { [ { } ]. [:a | { a } ]. [:a :b | { a. b } ]. [:a :b :c | { a. b. c } ] }.
blocks collect: [:b | b valueWithPossibleArgs: {1}]. blocks collect: [:b | b valueWithPossibleArgs: {1. 2. 3}]. blocks collect: [:b | b valueWithEnoughArguments: {1}]. blocks collect: [:b | b valueWithEnoughArguments: {1. 2. 3}].
I was also wondering how they relate to #cull:*
One of the major differences seems to be that valueWithP* and valueWithE* are both injecting nil for absent arguments, while normal #value* and #cull* methods signal an error. Is there a specific use case why one wouldnât want to be strict here as well, but instead inject nils?
Any comments or pointer appreciated.
Thanks Stefan
-- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
On 22 Feb 2018, at 10:41, Clément Bera <bera.clement@gmail.com> wrote:
On Thu, Feb 22, 2018 at 7:52 AM, Stephane Ducasse <stepharo.self@gmail.com> wrote: Clement
can you open a bug entry so that we clean this situation?
What is the bug?
Having two selectors doing the same thing is bad and confuses everybody. So that is a 'bug'.
You want to merge the two selectors and break compatibility with frameworks using the removed one?
Either we keep on adding cruft (or leaving cruft in place) in the sake of backwards compatibility or be do something about it. It is not hard to fix a couple of senders. It could be done more softly with a deprecation. Cleanups at the level of Object are particularly valuable.
Stef
On Mon, Feb 19, 2018 at 12:43 PM, Clément Bera <bera.clement@gmail.com> wrote:
Hi,
It seems the two methods have exactly the same behavior indeed. valueWithPossibleArgs: anArray valueWithEnoughArguments: anArray
One was edited recently but I think it's only to change the comment, they're both very old. My guess is that there are two for compatibility purpose (One is the selector that is considered as the most relevant that should be used, the other one is the one also present in other Smalltalks so we have it for cross-Smalltalk librairies or something like that), but only one is really needed. If you need only the concept for SOM-NS you can just implement one, if you want to be compatible with different Smalltalk lib implement both.
All use-cases of these methods I have found do not inject nils, they expect the block to have a number of arguments of the block less or equal to the number of parameters in the argument array. I would say they're used as #cullWithArguments: but for some reason other selector names were preferred.
Now, as you mentioned, these two methods are more than just cullWithArguments: since they inject nils if there are not enough parameters. To me it looks incorrect to do so because then while debugging your code you will get issues due to those injected nils and it will be tedious for the application programmer to track the problem down to these two methods.
There a few use-cases for nil injection though. Typically when changing existing frameworks in multiple repositories, it may be that during the update process the change to the caller is installed before the change of the callee, and if the code is actually used (code in UI for instance), injecting nils might avoid system break-down. Another use-case is for compatibility with frameworks using the nil injection, but I can't find a framework doing that right now.
Honestly, I would not implement the nil injection, but maybe some one else has a different point of view.
On Fri, Feb 16, 2018 at 6:25 PM, Stefan Marr <smalltalk@stefan-marr.de> wrote:
Hi:
I am trying to understand the different between and perhaps origin of BlockClosure>>#valueWithPossibleArgs: and BlockClosure>>#valueWithEnoughArguments:
I am trying to decide which of the two I need for SOMns.
The first one has seen more recent changes, when looking at the Pharo 6.1 download:
valueWithPossibleArgs: anArray â> 2/12/2017 StephaneDucasse valueWithEnoughArguments: anArray â> 3/11/2001 nk
While they have rather different implementations, they seem to behave identically, as far as I could tell using the following example:
blocks := { [ { } ]. [:a | { a } ]. [:a :b | { a. b } ]. [:a :b :c | { a. b. c } ] }.
blocks collect: [:b | b valueWithPossibleArgs: {1}]. blocks collect: [:b | b valueWithPossibleArgs: {1. 2. 3}]. blocks collect: [:b | b valueWithEnoughArguments: {1}]. blocks collect: [:b | b valueWithEnoughArguments: {1. 2. 3}].
I was also wondering how they relate to #cull:*
One of the major differences seems to be that valueWithP* and valueWithE* are both injecting nil for absent arguments, while normal #value* and #cull* methods signal an error. Is there a specific use case why one wouldnât want to be strict here as well, but instead inject nils?
Any comments or pointer appreciated.
Thanks Stefan
-- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
Thanks Sven I thought that suddenly I was an idiot. On Thu, Feb 22, 2018 at 11:47 AM, Sven Van Caekenberghe <sven@stfx.eu> wrote:
On 22 Feb 2018, at 10:41, Clément Bera <bera.clement@gmail.com> wrote:
On Thu, Feb 22, 2018 at 7:52 AM, Stephane Ducasse <stepharo.self@gmail.com> wrote: Clement
can you open a bug entry so that we clean this situation?
What is the bug?
Having two selectors doing the same thing is bad and confuses everybody. So that is a 'bug'.
You want to merge the two selectors and break compatibility with frameworks using the removed one?
Either we keep on adding cruft (or leaving cruft in place) in the sake of backwards compatibility or be do something about it. It is not hard to fix a couple of senders. It could be done more softly with a deprecation.
Cleanups at the level of Object are particularly valuable.
Stef
On Mon, Feb 19, 2018 at 12:43 PM, Clément Bera <bera.clement@gmail.com> wrote:
Hi,
It seems the two methods have exactly the same behavior indeed. valueWithPossibleArgs: anArray valueWithEnoughArguments: anArray
One was edited recently but I think it's only to change the comment, they're both very old. My guess is that there are two for compatibility purpose (One is the selector that is considered as the most relevant that should be used, the other one is the one also present in other Smalltalks so we have it for cross-Smalltalk librairies or something like that), but only one is really needed. If you need only the concept for SOM-NS you can just implement one, if you want to be compatible with different Smalltalk lib implement both.
All use-cases of these methods I have found do not inject nils, they expect the block to have a number of arguments of the block less or equal to the number of parameters in the argument array. I would say they're used as #cullWithArguments: but for some reason other selector names were preferred.
Now, as you mentioned, these two methods are more than just cullWithArguments: since they inject nils if there are not enough parameters. To me it looks incorrect to do so because then while debugging your code you will get issues due to those injected nils and it will be tedious for the application programmer to track the problem down to these two methods.
There a few use-cases for nil injection though. Typically when changing existing frameworks in multiple repositories, it may be that during the update process the change to the caller is installed before the change of the callee, and if the code is actually used (code in UI for instance), injecting nils might avoid system break-down. Another use-case is for compatibility with frameworks using the nil injection, but I can't find a framework doing that right now.
Honestly, I would not implement the nil injection, but maybe some one else has a different point of view.
On Fri, Feb 16, 2018 at 6:25 PM, Stefan Marr <smalltalk@stefan-marr.de> wrote:
Hi:
I am trying to understand the different between and perhaps origin of BlockClosure>>#valueWithPossibleArgs: and BlockClosure>>#valueWithEnoughArguments:
I am trying to decide which of the two I need for SOMns.
The first one has seen more recent changes, when looking at the Pharo 6.1 download:
valueWithPossibleArgs: anArray â> 2/12/2017 StephaneDucasse valueWithEnoughArguments: anArray â> 3/11/2001 nk
While they have rather different implementations, they seem to behave identically, as far as I could tell using the following example:
blocks := { [ { } ]. [:a | { a } ]. [:a :b | { a. b } ]. [:a :b :c | { a. b. c } ] }.
blocks collect: [:b | b valueWithPossibleArgs: {1}]. blocks collect: [:b | b valueWithPossibleArgs: {1. 2. 3}]. blocks collect: [:b | b valueWithEnoughArguments: {1}]. blocks collect: [:b | b valueWithEnoughArguments: {1. 2. 3}].
I was also wondering how they relate to #cull:*
One of the major differences seems to be that valueWithP* and valueWithE* are both injecting nil for absent arguments, while normal #value* and #cull* methods signal an error. Is there a specific use case why one wouldnât want to be strict here as well, but instead inject nils?
Any comments or pointer appreciated.
Thanks Stefan
-- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
++++++999^1000 :) On Fri, Feb 23, 2018 at 7:32 AM, Stephane Ducasse <stepharo.self@gmail.com> wrote:
Thanks Sven I thought that suddenly I was an idiot.
On Thu, Feb 22, 2018 at 11:47 AM, Sven Van Caekenberghe <sven@stfx.eu> wrote:
On 22 Feb 2018, at 10:41, Clément Bera <bera.clement@gmail.com> wrote:
On Thu, Feb 22, 2018 at 7:52 AM, Stephane Ducasse <
stepharo.self@gmail.com> wrote:
Clement
can you open a bug entry so that we clean this situation?
What is the bug?
Having two selectors doing the same thing is bad and confuses everybody. So that is a 'bug'.
You want to merge the two selectors and break compatibility with frameworks using the removed one?
Either we keep on adding cruft (or leaving cruft in place) in the sake of backwards compatibility or be do something about it. It is not hard to fix a couple of senders. It could be done more softly with a deprecation.
Cleanups at the level of Object are particularly valuable.
Stef
On Mon, Feb 19, 2018 at 12:43 PM, Clément Bera <bera.clement@gmail.com> wrote:
Hi,
It seems the two methods have exactly the same behavior indeed. valueWithPossibleArgs: anArray valueWithEnoughArguments: anArray
One was edited recently but I think it's only to change the comment, they're both very old. My guess is that there are two for compatibility purpose (One is the selector that is considered as the most relevant that should be used, the other one is the one also present in other Smalltalks so we have it for cross-Smalltalk librairies or something like that), but only one is really needed. If you need only the concept for SOM-NS you can just implement one, if you want to be compatible with different Smalltalk lib implement both.
All use-cases of these methods I have found do not inject nils, they expect the block to have a number of arguments of the block less or equal to the number of parameters in the argument array. I would say they're used as #cullWithArguments: but for some reason other selector names were preferred.
Now, as you mentioned, these two methods are more than just cullWithArguments: since they inject nils if there are not enough parameters. To me it looks incorrect to do so because then while debugging your code you will get issues due to those injected nils and it will be tedious for the application programmer to track the problem down to these two methods.
There a few use-cases for nil injection though. Typically when changing existing frameworks in multiple repositories, it may be that during the update process the change to the caller is installed before the change of the callee, and if the code is actually used (code in UI for instance), injecting nils might avoid system break-down. Another use-case is for compatibility with frameworks using the nil injection, but I can't find a framework doing that right now.
Honestly, I would not implement the nil injection, but maybe some one else has a different point of view.
On Fri, Feb 16, 2018 at 6:25 PM, Stefan Marr < smalltalk@stefan-marr.de> wrote:
Hi:
I am trying to understand the different between and perhaps origin of BlockClosure>>#valueWithPossibleArgs: and BlockClosure>>#valueWithEnoughArguments:
I am trying to decide which of the two I need for SOMns.
The first one has seen more recent changes, when looking at the
Pharo 6.1
download:
valueWithPossibleArgs: anArray â> 2/12/2017 StephaneDucasse valueWithEnoughArguments: anArray â> 3/11/2001 nk
While they have rather different implementations, they seem to behave identically, as far as I could tell using the following example:
blocks := { [ { } ]. [:a | { a } ]. [:a :b | { a. b } ]. [:a :b :c | { a. b. c } ] }.
blocks collect: [:b | b valueWithPossibleArgs: {1}]. blocks collect: [:b | b valueWithPossibleArgs: {1. 2. 3}]. blocks collect: [:b | b valueWithEnoughArguments: {1}]. blocks collect: [:b | b valueWithEnoughArguments: {1. 2. 3}].
I was also wondering how they relate to #cull:*
One of the major differences seems to be that valueWithP* and valueWithE* are both injecting nil for absent arguments, while normal #value* and #cull* methods signal an error. Is there a specific use case why one wouldnât want to be strict here as well, but instead inject nils?
Any comments or pointer appreciated.
Thanks Stefan
-- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
-- Guille Polito Research Engineer Centre de Recherche en Informatique, Signal et Automatique de Lille CRIStAL - UMR 9189 French National Center for Scientific Research - *http://www.cnrs.fr <http://www.cnrs.fr>* *Web:* *http://guillep.github.io* <http://guillep.github.io> *Phone: *+33 06 52 70 66 13
On Thu, Feb 22, 2018 at 11:47 AM, Sven Van Caekenberghe <sven@stfx.eu> wrote:
On 22 Feb 2018, at 10:41, Clément Bera <bera.clement@gmail.com> wrote:
On Thu, Feb 22, 2018 at 7:52 AM, Stephane Ducasse < stepharo.self@gmail.com> wrote: Clement
can you open a bug entry so that we clean this situation?
What is the bug?
Having two selectors doing the same thing is bad and confuses everybody. So that is a 'bug'.
You want to merge the two selectors and break compatibility with frameworks using the removed one?
Either we keep on adding cruft (or leaving cruft in place) in the sake of backwards compatibility or be do something about it. It is not hard to fix a couple of senders. It could be done more softly with a deprecation.
Cleanups at the level of Object are particularly valuable.
Ok. Is the nil injection a feature or a bug too ? Should it be removed ?
Stef
On Mon, Feb 19, 2018 at 12:43 PM, Clément Bera <bera.clement@gmail.com> wrote:
Hi,
It seems the two methods have exactly the same behavior indeed. valueWithPossibleArgs: anArray valueWithEnoughArguments: anArray
One was edited recently but I think it's only to change the comment, they're both very old. My guess is that there are two for compatibility purpose (One is the selector that is considered as the most relevant that should be used, the other one is the one also present in other Smalltalks so we have it for cross-Smalltalk librairies or something like that), but only one is really needed. If you need only the concept for SOM-NS you can just implement one, if you want to be compatible with different Smalltalk lib implement both.
All use-cases of these methods I have found do not inject nils, they expect the block to have a number of arguments of the block less or equal to the number of parameters in the argument array. I would say they're used as #cullWithArguments: but for some reason other selector names were preferred.
Now, as you mentioned, these two methods are more than just cullWithArguments: since they inject nils if there are not enough parameters. To me it looks incorrect to do so because then while debugging your code you will get issues due to those injected nils and it will be tedious for the application programmer to track the problem down to these two methods.
There a few use-cases for nil injection though. Typically when changing existing frameworks in multiple repositories, it may be that during the update process the change to the caller is installed before the change of the callee, and if the code is actually used (code in UI for instance), injecting nils might avoid system break-down. Another use-case is for compatibility with frameworks using the nil injection, but I can't find a framework doing that right now.
Honestly, I would not implement the nil injection, but maybe some one else has a different point of view.
On Fri, Feb 16, 2018 at 6:25 PM, Stefan Marr <smalltalk@stefan-marr.de
wrote:
Hi:
I am trying to understand the different between and perhaps origin of BlockClosure>>#valueWithPossibleArgs: and BlockClosure>>#valueWithEnoughArguments:
I am trying to decide which of the two I need for SOMns.
The first one has seen more recent changes, when looking at the Pharo
6.1
download:
valueWithPossibleArgs: anArray â> 2/12/2017 StephaneDucasse valueWithEnoughArguments: anArray â> 3/11/2001 nk
While they have rather different implementations, they seem to behave identically, as far as I could tell using the following example:
blocks := { [ { } ]. [:a | { a } ]. [:a :b | { a. b } ]. [:a :b :c | { a. b. c } ] }.
blocks collect: [:b | b valueWithPossibleArgs: {1}]. blocks collect: [:b | b valueWithPossibleArgs: {1. 2. 3}]. blocks collect: [:b | b valueWithEnoughArguments: {1}]. blocks collect: [:b | b valueWithEnoughArguments: {1. 2. 3}].
I was also wondering how they relate to #cull:*
One of the major differences seems to be that valueWithP* and valueWithE* are both injecting nil for absent arguments, while normal #value* and #cull* methods signal an error. Is there a specific use case why one wouldnât want to be strict here as well, but instead inject nils?
Any comments or pointer appreciated.
Thanks Stefan
-- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
-- Clément Béra Pharo consortium engineer https://clementbera.wordpress.com/ Bâtiment B 40, avenue Halley 59650 Villeneuve d'Ascq
Sven Van Caekenberghe <sven@stfx.eu> wrote:
Either we keep on adding cruft (or leaving cruft in place) in the sake of backwards compatibility or be do something about it. It is not hard to fix a couple of senders. It could be done more softly with a deprecation.
Cleanups at the level of Object are particularly valuable.
It needs to be done in a way that is sustainable. That means with a code rewrite rule. Deprecations are not enough. The number of people who can migrate forwards older code is very small, especially as we are missing crucial parts of design discussion that took place on slack and were not copied to mail. When I created my Morphic screencasts I had to look back to mailing list discussions from 10 years ago and had to run ancient squeak images to gain an understanding of how things were supposed to work and how they changed. Cleanups at this low level need to be done with rewrite rules (and tests) Stephan
2018-02-23 14:50 GMT-03:00 Stephan Eggermont <stephan@stack.nl>:
Sven Van Caekenberghe <sven@stfx.eu> wrote:
Cleanups at the level of Object are particularly valuable.
It needs to be done in a way that is sustainable. That means with a code rewrite rule. Deprecations are not enough. The number of people who can migrate forwards older code is very small, especially as we are missing crucial parts of design discussion that took place on slack and were not copied to mail.
Why not both? With a noisy enough deprecation warning and a handy rewrite rules script it would ease the transition. AFAIR the procedure was to mark as deprecated in one release and remove completely in the next one. Best regards! Esteban A. Maringolo
Esteban A. Maringolo wrote
That means with a code rewrite rule. Deprecations are not enough. Why not both?
It seems like you're both saying the same thing, no? Isn't this all accomplished via: self deprecated: 'Renamed to #action because now accepts any valuable as an action (e.g. aBlock).' on: '10/24/2017' in: #Pharo61 transformWith: '`@receiver selector' -> '`@receiver action'. Or are you saying there should be a collection of rewrite rules even after the method is totally removed so one can bring a very old project up to the newest APIs? ----- Cheers, Sean -- Sent from: http://forum.world.st/Pharo-Smalltalk-Developers-f1294837.html
Sean P. DeNigris <sean@clipperadams.com> wrote:
Or are you saying there should be a collection of rewrite rules even after the method is totally removed so one can bring a very old project up to the newest APIs?
Making sure the code works in each intermediate version of Pharo is increasingly unattractive. It also requires indefinite support of all the old pharo versions. A collection of separate rewrite rules looks to me like the right thing to both explain changes and speed up migrations. Stephan
Stephan Can you stop ranting for ranting and be positive thinking? Don't you think that are ARE PAYING ATTENTION!!! We will NOT END UP looking like some other systems! If you want to help here is a concrete proposal (we want to do it but we are busy doing bullshit works like reviewing fix and fixing typos around). - Gather all the deprecated methods in the past Pharo versions starting with Pharo 60. - add Transform rule when posible - package them so that people can load it and get helped during their migration EASY NO? sorry but I have something to hear that such kind of emails. Stef On Fri, Feb 23, 2018 at 8:20 PM, Stephan Eggermont <stephan@stack.nl> wrote:
Sean P. DeNigris <sean@clipperadams.com> wrote:
Or are you saying there should be a collection of rewrite rules even after the method is totally removed so one can bring a very old project up to the newest APIs?
Making sure the code works in each intermediate version of Pharo is increasingly unattractive. It also requires indefinite support of all the old pharo versions. A collection of separate rewrite rules looks to me like the right thing to both explain changes and speed up migrations.
Stephan
On Fri, Feb 16, 2018 at 2:25 PM, Stefan Marr <smalltalk@stefan-marr.de> wrote:
Hi:
I am trying to understand the different between and perhaps origin of BlockClosure>>#valueWithPossibleArgs: and BlockClosure>># valueWithEnoughArguments:
I am trying to decide which of the two I need for SOMns.
The first one has seen more recent changes, when looking at the Pharo 6.1 download:
valueWithPossibleArgs: anArray â> 2/12/2017 StephaneDucasse valueWithEnoughArguments: anArray â> 3/11/2001 nk
While they have rather different implementations, they seem to behave identically, as far as I could tell using the following example:
blocks := { [ { } ]. [:a | { a } ]. [:a :b | { a. b } ]. [:a :b :c | { a. b. c } ] }.
blocks collect: [:b | b valueWithPossibleArgs: {1}]. blocks collect: [:b | b valueWithPossibleArgs: {1. 2. 3}]. blocks collect: [:b | b valueWithEnoughArguments: {1}]. blocks collect: [:b | b valueWithEnoughArguments: {1. 2. 3}].
I was also wondering how they relate to #cull:*
One of the major differences seems to be that valueWithP* and valueWithE* are both injecting nil for absent arguments, while normal #value* and #cull* methods signal an error. Is there a specific use case why one wouldnât want to be strict here as well, but instead inject nils?
I guess it's because we offer both flavors. If you see valueWithArguments: you will see it does throw errors like #value* / #cull* do.
Any comments or pointer appreciated.
Thanks Stefan
-- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
-- Mariano http://marianopeck.wordpress.com
Hi Mariano:
On 19 Feb 2018, at 20:09, Mariano Martinez Peck <marianopeck@gmail.com> wrote:
I guess it's because we offer both flavors. If you see valueWithArguments: you will see it does throw errors like #value* / #cull* do.
Did you try my examples? I donât think this is correct. In Pharo 6.1, #valueWithArguments: doesnât throw any errors as far as I can see. It inserts nil for absent arguments however. This is what I tried: blocks := { [ { } ]. [:a | { a } ]. [:a :b | { a. b } ]. [:a :b :c | { a. b. c } ] }. blocks collect: [:b | b valueWithPossibleArgs: {1}]. blocks collect: [:b | b valueWithPossibleArgs: {1. 2. 3}]. blocks collect: [:b | b valueWithEnoughArguments: {1}]. blocks collect: [:b | b valueWithEnoughArguments: {1. 2. 3}]. Best regards Stefan -- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
Ah, sorry, yes #valueWithArguments: does throw errors. Best regards Stefan
On 19 Feb 2018, at 20:17, Stefan Marr <smalltalk@stefan-marr.de> wrote:
Hi Mariano:
On 19 Feb 2018, at 20:09, Mariano Martinez Peck <marianopeck@gmail.com> wrote:
I guess it's because we offer both flavors. If you see valueWithArguments: you will see it does throw errors like #value* / #cull* do.
Did you try my examples? I donât think this is correct. In Pharo 6.1, #valueWithArguments: doesnât throw any errors as far as I can see. It inserts nil for absent arguments however.
This is what I tried:
blocks := { [ { } ]. [:a | { a } ]. [:a :b | { a. b } ]. [:a :b :c | { a. b. c } ] }.
blocks collect: [:b | b valueWithPossibleArgs: {1}]. blocks collect: [:b | b valueWithPossibleArgs: {1. 2. 3}]. blocks collect: [:b | b valueWithEnoughArguments: {1}]. blocks collect: [:b | b valueWithEnoughArguments: {1. 2. 3}].
Best regards Stefan
-- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
-- Stefan Marr School of Computing, University of Kent http://stefan-marr.de/research/
participants (9)
-
Clément Bera -
Esteban A. Maringolo -
Guillermo Polito -
Mariano Martinez Peck -
Sean P. DeNigris -
Stefan Marr -
Stephan Eggermont -
Stephane Ducasse -
Sven Van Caekenberghe