TextAttributes in source code or class comment?
Hi all, The source code of a method could be a Text, with TextAttributes. In such case, it will be written down to a stream with a "]style[" suffix. Through a special parsing, the system can reconstruct the original Text. I think the use of Text in source code or class comments is discouraged, but I understand we kept support for some backward compatibility. Now, may be it's time to remove this feature. With Pablo, we looked for uses of ]style[ in the .sources and .changes files of a new Pharo 5.0, and there are no real uses. It appears in some comments only, as examples. We tried to find past discussions about this feature, because we're sure there were, but we couldn't. So, the questions is: Can we remove support for text in source code or class comments in Pharo 5.0? I think I can propose a slice. Regards, MartÃn PS: I was trying to cut dependencies of Text-Core with System-Changes and System-Changes.
Hi Martin, On Fri, Sep 4, 2015 at 8:13 AM, Martin Dias <tinchodias@gmail.com> wrote:
Hi all,
The source code of a method could be a Text, with TextAttributes. In such case, it will be written down to a stream with a "]style[" suffix. Through a special parsing, the system can reconstruct the original Text. I think the use of Text in source code or class comments is discouraged, but I understand we kept support for some backward compatibility. Now, may be it's time to remove this feature.
With Pablo, we looked for uses of ]style[ in the .sources and .changes files of a new Pharo 5.0, and there are no real uses. It appears in some comments only, as examples.
We tried to find past discussions about this feature, because we're sure there were, but we couldn't.
So, the questions is:
Can we remove support for text in source code or class comments in Pharo 5.0?
IIUC, ]style[ also supports things like embedding URL links and method references in class comments. So I would request that it /not/ be removed. It is very useful to put clickable links in class comments.
I think I can propose a slice.
Regards, MartÃn PS: I was trying to cut dependencies of Text-Core with System-Changes and System-Changes.
-- _,,,^..^,,,_ best, Eliot
eliot Nobody defines urls in method comment anymore. In addition would prefer to have an explicit syntax as in Pillar or html to represent an hyperlink. We could have *Class>>#method* when pillar is not loaded and else a nicer representation. ctrl-click on a class should jump on it already and we could map it to *Class>>#method* In class comments we should use pillar and get a real text renderer.
Hi Martin,
On Fri, Sep 4, 2015 at 8:13 AM, Martin Dias <tinchodias@gmail.com <mailto:tinchodias@gmail.com>> wrote:
Hi all,
The source code of a method could be a Text, with TextAttributes. In such case, it will be written down to a stream with a "]style[" suffix. Through a special parsing, the system can reconstruct the original Text. I think the use of Text in source code or class comments is discouraged, but I understand we kept support for some backward compatibility. Now, may be it's time to remove this feature.
With Pablo, we looked for uses of ]style[ in the .sources and .changes files of a new Pharo 5.0, and there are no real uses. It appears in some comments only, as examples.
We tried to find past discussions about this feature, because we're sure there were, but we couldn't.
So, the questions is:
Can we remove support for text in source code or class comments in Pharo 5.0?
IIUC, ]style[ also supports things like embedding URL links and method references in class comments. So I would request that it /not/ be removed. It is very useful to put clickable links in class comments.
I think I can propose a slice.
Regards, MartÃn PS: I was trying to cut dependencies of Text-Core with System-Changes and System-Changes.
-- _,,,^..^,,,_ best, Eliot
Sent from my iPhone
On Sep 4, 2015, at 1:33 PM, stepharo <stepharo@free.fr> wrote:
eliot
Nobody defines urls in method comment anymore.
I do. Several places in the Cog JIT contain irks to specification documents. I don't care how its implemented (providing it's preserved in text, or maybe have the pretty-printer scan comments for http://. But it is sad that we can't have hyperlinks in text. Look I know the implementation stinks, but it can be fixed. But if the functionality is lost the system is weakened. Instead of asking what you can get rid if why not ask what you can reimplement or enhance more elegantly. Christ you're even using minis paced fonts. It's like the dark ages in there ;-)
In addition would prefer to have an explicit syntax as in Pillar or html to represent an hyperlink. We could have *Class>>#method* when pillar is not loaded and else a nicer representation. ctrl-click on a class should jump on it already and we could map it to *Class>>#method*
In class comments we should use pillar and get a real text renderer.
Hi Martin,
On Fri, Sep 4, 2015 at 8:13 AM, Martin Dias <tinchodias@gmail.com> wrote: Hi all,
The source code of a method could be a Text, with TextAttributes. In such case, it will be written down to a stream with a "]style[" suffix. Through a special parsing, the system can reconstruct the original Text. I think the use of Text in source code or class comments is discouraged, but I understand we kept support for some backward compatibility. Now, may be it's time to remove this feature.
With Pablo, we looked for uses of ]style[ in the .sources and .changes files of a new Pharo 5.0, and there are no real uses. It appears in some comments only, as examples.
We tried to find past discussions about this feature, because we're sure there were, but we couldn't.
So, the questions is:
Can we remove support for text in source code or class comments in Pharo 5.0?
IIUC, ]style[ also supports things like embedding URL links and method references in class comments. So I would request that it /not/ be removed. It is very useful to put clickable links in class comments.
I think I can propose a slice.
Regards, MartÃn PS: I was trying to cut dependencies of Text-Core with System-Changes and System-Changes.
-- _,,,^..^,,,_ best, Eliot
On Sat, Sep 5, 2015 at 4:43 AM, Eliot Miranda <eliot.miranda@gmail.com> wrote:
On Sep 4, 2015, at 1:33 PM, stepharo <stepharo@free.fr> wrote:
Nobody defines urls in method comment anymore.
I do. Several places in the Cog JIT contain irks to specification documents. I don't care how its implemented (providing it's preserved in text, or maybe have the pretty-printer scan comments for http://. But it is sad that we can't have hyperlinks in text.
Look I know the implementation stinks, but it can be fixed. But if the functionality is lost the system is weakened. Instead of asking what you can get rid if why not ask what you can reimplement or enhance more elegantly.
Its a reasonable strategy to start the discussion at the extreme end and work back from there :) But if its being used we should think harder about it. What alternatives/workarounds are there? First just to clarify, the aim is to remove Text from the image for the boot-strap? 1. Save the text internally as a string in the stream format and up-convert to Text only when its being displayed? But I guess these styles aren't restricted to just occurring in comments and maybe cause problem for compilation ?? 2. As a stepping stone, introduce a TextPlaceholder class within minimal interface required by compilation which simply strips out all styles when Text is missing from the Image. When Text is present it #becomes back to it. Maybe risks loosing styles editing is done on the raw-stream-format. Maybe any editor would show the raw ]style[ and the compiler asks for #source which strips styles out. 3. Introduce Pillar and convert across. What are the requirements? a. URLs b. Class/method links c. Would you expect that Pillar can only appear in comments? Or in-line with code? d. Colouring? ...? btw, I haven't come across this before. Can someone point me to some url & styled code where I can have a look at it cheers -ben
Christ you're even using minis paced fonts. It's like the dark ages in there ;-)
In addition would prefer to have an explicit syntax as in Pillar or html to represent an hyperlink. We could have *Class>>#method* when pillar is not loaded and else a nicer representation. ctrl-click on a class should jump on it already and we could map it to *Class>>#method*
In class comments we should use pillar and get a real text renderer.
Hi Martin,
On Fri, Sep 4, 2015 at 8:13 AM, Martin Dias <tinchodias@gmail.com> wrote:
Hi all,
The source code of a method could be a Text, with TextAttributes. In such case, it will be written down to a stream with a "]style[" suffix. Through a special parsing, the system can reconstruct the original Text. I think the use of Text in source code or class comments is discouraged, but I understand we kept support for some backward compatibility. Now, may be it's time to remove this feature.
With Pablo, we looked for uses of ]style[ in the .sources and .changes files of a new Pharo 5.0, and there are no real uses. It appears in some comments only, as examples.
We tried to find past discussions about this feature, because we're sure there were, but we couldn't.
So, the questions is:
Can we remove support for text in source code or class comments in Pharo 5.0?
IIUC, ]style[ also supports things like embedding URL links and method references in class comments. So I would request that it /not/ be removed. It is very useful to put clickable links in class comments.
I think I can propose a slice.
Regards, MartÃn PS: I was trying to cut dependencies of Text-Core with System-Changes and System-Changes.
-- _,,,^..^,,,_ best, Eliot
Hi all On Sat, Sep 5, 2015 at 3:44 AM, Ben Coman <btc@openinworld.com> wrote:
On Sat, Sep 5, 2015 at 4:43 AM, Eliot Miranda <eliot.miranda@gmail.com> wrote:
On Sep 4, 2015, at 1:33 PM, stepharo <stepharo@free.fr> wrote:
Nobody defines urls in method comment anymore.
I do. Several places in the Cog JIT contain irks to specification documents. I don't care how its implemented (providing it's preserved in text, or maybe have the pretty-printer scan comments for http://. But it is sad that we can't have hyperlinks in text.
Look I know the implementation stinks, but it can be fixed. But if the functionality is lost the system is weakened. Instead of asking what you can get rid if why not ask what you can reimplement or enhance more elegantly.
Its a reasonable strategy to start the discussion at the extreme end and work back from there :) But if its being used we should think harder about it. What alternatives/workarounds are there?
First just to clarify, the aim is to remove Text from the image for the boot-strap?
Yes :) I heard that we want to do cut the dependency to Text-Core, so I browsed how to do it. Ben, there is an explanation about ]style[ in the class comment of RemoteString. Eliot, I didn't thought that somebody was using Text. In fact, I don't know how to do it in Pharo... how do you do? I found that ChangeRecord>>fileIn uses #newTextChunk to parse, but I just tested filing-in the attached .st file and it doesn't work. I guess the new CodeImporter replaced the ChangeRecord implementation, and it raises an error when ]style[. MartÃn
1. Save the text internally as a string in the stream format and up-convert to Text only when its being displayed? But I guess these styles aren't restricted to just occurring in comments and maybe cause problem for compilation ??
2. As a stepping stone, introduce a TextPlaceholder class within minimal interface required by compilation which simply strips out all styles when Text is missing from the Image. When Text is present it #becomes back to it. Maybe risks loosing styles editing is done on the raw-stream-format. Maybe any editor would show the raw ]style[ and the compiler asks for #source which strips styles out.
3. Introduce Pillar and convert across. What are the requirements? a. URLs b. Class/method links c. Would you expect that Pillar can only appear in comments? Or in-line with code? d. Colouring? ...?
btw, I haven't come across this before. Can someone point me to some url & styled code where I can have a look at it
cheers -ben
Christ you're even using minis paced fonts. It's like the dark ages in there ;-)
In addition would prefer to have an explicit syntax as in Pillar or html to represent an hyperlink. We could have *Class>>#method* when pillar is not loaded and else a nicer representation. ctrl-click on a class should jump on it already and we could map it to *Class>>#method*
In class comments we should use pillar and get a real text renderer.
Hi Martin,
On Fri, Sep 4, 2015 at 8:13 AM, Martin Dias <tinchodias@gmail.com> wrote:
Hi all,
The source code of a method could be a Text, with TextAttributes. In such case, it will be written down to a stream with a "]style[" suffix. Through a special parsing, the system can reconstruct the original Text. I think the use of Text in source code or class comments is discouraged, but I understand we kept support for some backward compatibility. Now, may be it's time to remove this feature.
With Pablo, we looked for uses of ]style[ in the .sources and .changes files of a new Pharo 5.0, and there are no real uses. It appears in some comments only, as examples.
We tried to find past discussions about this feature, because we're sure there were, but we couldn't.
So, the questions is:
Can we remove support for text in source code or class comments in Pharo 5.0?
IIUC, ]style[ also supports things like embedding URL links and method references in class comments. So I would request that it /not/ be removed. It is very useful to put clickable links in class comments.
I think I can propose a slice.
Regards, MartÃn PS: I was trying to cut dependencies of Text-Core with System-Changes and System-Changes.
-- _,,,^..^,,,_ best, Eliot
The same for the ]lang[ tag (see issue 15750) 2015-09-05 12:06 GMT+02:00 Martin Dias <tinchodias@gmail.com>:
Hi all
On Sat, Sep 5, 2015 at 3:44 AM, Ben Coman <btc@openinworld.com> wrote:
On Sat, Sep 5, 2015 at 4:43 AM, Eliot Miranda <eliot.miranda@gmail.com> wrote:
On Sep 4, 2015, at 1:33 PM, stepharo <stepharo@free.fr> wrote:
Nobody defines urls in method comment anymore.
I do. Several places in the Cog JIT contain irks to specification documents. I don't care how its implemented (providing it's preserved
in
text, or maybe have the pretty-printer scan comments for http://. But it is sad that we can't have hyperlinks in text.
Look I know the implementation stinks, but it can be fixed. But if the functionality is lost the system is weakened. Instead of asking what you can get rid if why not ask what you can reimplement or enhance more elegantly.
Its a reasonable strategy to start the discussion at the extreme end and work back from there :) But if its being used we should think harder about it. What alternatives/workarounds are there?
First just to clarify, the aim is to remove Text from the image for the boot-strap?
Yes :) I heard that we want to do cut the dependency to Text-Core, so I browsed how to do it.
Ben, there is an explanation about ]style[ in the class comment of RemoteString.
Eliot, I didn't thought that somebody was using Text. In fact, I don't know how to do it in Pharo... how do you do? I found that ChangeRecord>>fileIn uses #newTextChunk to parse, but I just tested filing-in the attached .st file and it doesn't work. I guess the new CodeImporter replaced the ChangeRecord implementation, and it raises an error when ]style[.
MartÃn
1. Save the text internally as a string in the stream format and up-convert to Text only when its being displayed? But I guess these styles aren't restricted to just occurring in comments and maybe cause problem for compilation ??
2. As a stepping stone, introduce a TextPlaceholder class within minimal interface required by compilation which simply strips out all styles when Text is missing from the Image. When Text is present it #becomes back to it. Maybe risks loosing styles editing is done on the raw-stream-format. Maybe any editor would show the raw ]style[ and the compiler asks for #source which strips styles out.
3. Introduce Pillar and convert across. What are the requirements? a. URLs b. Class/method links c. Would you expect that Pillar can only appear in comments? Or in-line with code? d. Colouring? ...?
btw, I haven't come across this before. Can someone point me to some url & styled code where I can have a look at it
cheers -ben
Christ you're even using minis paced fonts. It's like the dark ages in there ;-)
In addition would prefer to have an explicit syntax as in Pillar or
html to
represent an hyperlink. We could have *Class>>#method* when pillar is not loaded and else a nicer representation. ctrl-click on a class should jump on it already and we could map it to *Class>>#method*
In class comments we should use pillar and get a real text renderer.
Hi Martin,
On Fri, Sep 4, 2015 at 8:13 AM, Martin Dias <tinchodias@gmail.com> wrote:
Hi all,
The source code of a method could be a Text, with TextAttributes. In
such
case, it will be written down to a stream with a "]style[" suffix. Through a special parsing, the system can reconstruct the original Text. I think the use of Text in source code or class comments is discouraged, but I understand we kept support for some backward compatibility. Now, may be it's time to remove this feature.
With Pablo, we looked for uses of ]style[ in the .sources and .changes files of a new Pharo 5.0, and there are no real uses. It appears in some comments only, as examples.
We tried to find past discussions about this feature, because we're sure there were, but we couldn't.
So, the questions is:
Can we remove support for text in source code or class comments in Pharo 5.0?
IIUC, ]style[ also supports things like embedding URL links and method references in class comments. So I would request that it /not/ be removed. It is very useful to put clickable links in class comments.
I think I can propose a slice.
Regards, MartÃn PS: I was trying to cut dependencies of Text-Core with System-Changes
and
System-Changes.
-- _,,,^..^,,,_ best, Eliot
On Sat, Sep 5, 2015 at 1:10 PM, Nicolai Hess <nicolaihess@web.de> wrote:
The same for the ]lang[ tag (see issue 15750)
Ok. But I still don't know how to add a url to a method's source code or comment from the UI. I mean, people with a non-English keyboard can produce source code with ]lang[, but how can I produce a ]style[ ? MartÃn
2015-09-05 12:06 GMT+02:00 Martin Dias <tinchodias@gmail.com>:
Hi all
On Sat, Sep 5, 2015 at 3:44 AM, Ben Coman <btc@openinworld.com> wrote:
On Sat, Sep 5, 2015 at 4:43 AM, Eliot Miranda <eliot.miranda@gmail.com> wrote:
On Sep 4, 2015, at 1:33 PM, stepharo <stepharo@free.fr> wrote:
Nobody defines urls in method comment anymore.
I do. Several places in the Cog JIT contain irks to specification documents. I don't care how its implemented (providing it's preserved in text, or maybe have the pretty-printer scan comments for http://. But it is sad that we can't have hyperlinks in text.
Look I know the implementation stinks, but it can be fixed. But if the functionality is lost the system is weakened. Instead of asking what you can get rid if why not ask what you can reimplement or enhance more elegantly.
Its a reasonable strategy to start the discussion at the extreme end and work back from there :) But if its being used we should think harder about it. What alternatives/workarounds are there?
First just to clarify, the aim is to remove Text from the image for the boot-strap?
Yes :) I heard that we want to do cut the dependency to Text-Core, so I browsed how to do it.
Ben, there is an explanation about ]style[ in the class comment of RemoteString.
Eliot, I didn't thought that somebody was using Text. In fact, I don't know how to do it in Pharo... how do you do? I found that ChangeRecord>>fileIn uses #newTextChunk to parse, but I just tested filing-in the attached .st file and it doesn't work. I guess the new CodeImporter replaced the ChangeRecord implementation, and it raises an error when ]style[.
MartÃn
1. Save the text internally as a string in the stream format and up-convert to Text only when its being displayed? But I guess these styles aren't restricted to just occurring in comments and maybe cause problem for compilation ??
2. As a stepping stone, introduce a TextPlaceholder class within minimal interface required by compilation which simply strips out all styles when Text is missing from the Image. When Text is present it #becomes back to it. Maybe risks loosing styles editing is done on the raw-stream-format. Maybe any editor would show the raw ]style[ and the compiler asks for #source which strips styles out.
3. Introduce Pillar and convert across. What are the requirements? a. URLs b. Class/method links c. Would you expect that Pillar can only appear in comments? Or in-line with code? d. Colouring? ...?
btw, I haven't come across this before. Can someone point me to some url & styled code where I can have a look at it
cheers -ben
Christ you're even using minis paced fonts. It's like the dark ages in there ;-)
In addition would prefer to have an explicit syntax as in Pillar or html to represent an hyperlink. We could have *Class>>#method* when pillar is not loaded and else a nicer representation. ctrl-click on a class should jump on it already and we could map it to *Class>>#method*
In class comments we should use pillar and get a real text renderer.
Hi Martin,
On Fri, Sep 4, 2015 at 8:13 AM, Martin Dias <tinchodias@gmail.com> wrote:
Hi all,
The source code of a method could be a Text, with TextAttributes. In such case, it will be written down to a stream with a "]style[" suffix. Through a special parsing, the system can reconstruct the original Text. I think the use of Text in source code or class comments is discouraged, but I understand we kept support for some backward compatibility. Now, may be it's time to remove this feature.
With Pablo, we looked for uses of ]style[ in the .sources and .changes files of a new Pharo 5.0, and there are no real uses. It appears in some comments only, as examples.
We tried to find past discussions about this feature, because we're sure there were, but we couldn't.
So, the questions is:
Can we remove support for text in source code or class comments in Pharo 5.0?
IIUC, ]style[ also supports things like embedding URL links and method references in class comments. So I would request that it /not/ be removed. It is very useful to put clickable links in class comments.
I think I can propose a slice.
Regards, MartÃn PS: I was trying to cut dependencies of Text-Core with System-Changes and System-Changes.
-- _,,,^..^,,,_ best, Eliot
2015-09-08 11:03 GMT+02:00 Martin Dias <tinchodias@gmail.com>:
On Sat, Sep 5, 2015 at 1:10 PM, Nicolai Hess <nicolaihess@web.de> wrote:
The same for the ]lang[ tag (see issue 15750)
Ok. But I still don't know how to add a url to a method's source code or comment from the UI.
I mean, people with a non-English keyboard can produce source code with ]lang[, but how can I produce a ]style[ ?
Hi Martin, In squeak, you can open a "choose style" dialog (select some text and use shortcut cmd+6). This lets you choose a text color or text links. For example, this is working in the class comment editor too, you can open the comment for class Morph, select a word like "Rectangle" and put a link to that class definition. This style attributes are first only a part of the Text instance, but saving or file out the class definition will include this style attributes with the ]style[ tag. Writing the style code text is not supported in pharo anymore. Changing the style (color and text links) with cmd+6 does not work in Rubrics text components anymore. It is working with the old PluggableTextMorph (but only color attributes - no links). Reading Text with style attributes is still working in Pharo (only color and emphasis, no text links). For example, put this "styled" text in a Workspace and inspect it: 'Red Gree Blue! ]style[(3 1 4 1 4)f1cred;b,f1b,f1cgreen;b,f1b,f1cblue;b!' readStream nextChunkText
MartÃn
2015-09-05 12:06 GMT+02:00 Martin Dias <tinchodias@gmail.com>:
Hi all
On Sat, Sep 5, 2015 at 3:44 AM, Ben Coman <btc@openinworld.com> wrote:
On Sat, Sep 5, 2015 at 4:43 AM, Eliot Miranda <
eliot.miranda@gmail.com>
wrote:
On Sep 4, 2015, at 1:33 PM, stepharo <stepharo@free.fr> wrote:
Nobody defines urls in method comment anymore.
I do. Several places in the Cog JIT contain irks to specification documents. I don't care how its implemented (providing it's
preserved
in text, or maybe have the pretty-printer scan comments for http://. But it is sad that we can't have hyperlinks in text.
Look I know the implementation stinks, but it can be fixed. But if the functionality is lost the system is weakened. Instead of asking what you can get rid if why not ask what you can reimplement or enhance more elegantly.
Its a reasonable strategy to start the discussion at the extreme end and work back from there :) But if its being used we should think harder about it. What alternatives/workarounds are there?
First just to clarify, the aim is to remove Text from the image for the boot-strap?
Yes :) I heard that we want to do cut the dependency to Text-Core, so I browsed how to do it.
Ben, there is an explanation about ]style[ in the class comment of RemoteString.
Eliot, I didn't thought that somebody was using Text. In fact, I don't know how to do it in Pharo... how do you do? I found that ChangeRecord>>fileIn uses #newTextChunk to parse, but I just tested filing-in the attached .st file and it doesn't work. I guess the new CodeImporter replaced the ChangeRecord implementation, and it raises an error when ]style[.
MartÃn
1. Save the text internally as a string in the stream format and up-convert to Text only when its being displayed? But I guess these styles aren't restricted to just occurring in comments and maybe cause problem for compilation ??
2. As a stepping stone, introduce a TextPlaceholder class within minimal interface required by compilation which simply strips out all styles when Text is missing from the Image. When Text is present it #becomes back to it. Maybe risks loosing styles editing is done on the raw-stream-format. Maybe any editor would show the raw ]style[ and the compiler asks for #source which strips styles out.
3. Introduce Pillar and convert across. What are the requirements? a. URLs b. Class/method links c. Would you expect that Pillar can only appear in comments? Or in-line with code? d. Colouring? ...?
btw, I haven't come across this before. Can someone point me to some url & styled code where I can have a look at it
cheers -ben
Christ you're even using minis paced fonts. It's like the dark ages in there ;-)
In addition would prefer to have an explicit syntax as in Pillar or html to represent an hyperlink. We could have *Class>>#method* when pillar is not loaded and else a nicer representation. ctrl-click on a class should jump on it already and we could map it
to
*Class>>#method*
In class comments we should use pillar and get a real text renderer.
Hi Martin,
On Fri, Sep 4, 2015 at 8:13 AM, Martin Dias <tinchodias@gmail.com> wrote:
Hi all,
The source code of a method could be a Text, with TextAttributes. In such case, it will be written down to a stream with a "]style[" suffix. Through a special parsing, the system can reconstruct the original Text. I
think
the use of Text in source code or class comments is discouraged, but I understand we kept support for some backward compatibility. Now, may be it's time to remove this feature.
With Pablo, we looked for uses of ]style[ in the .sources and .changes files of a new Pharo 5.0, and there are no real uses. It appears in some comments only, as examples.
We tried to find past discussions about this feature, because we're sure there were, but we couldn't.
So, the questions is:
Can we remove support for text in source code or class comments in Pharo 5.0?
IIUC, ]style[ also supports things like embedding URL links and method references in class comments. So I would request that it /not/ be removed. It is very useful to put clickable links in class comments.
I think I can propose a slice.
Regards, MartÃn PS: I was trying to cut dependencies of Text-Core with
System-Changes
and System-Changes.
-- _,,,^..^,,,_ best, Eliot
On 08 Sep 2015, at 11:03 , Martin Dias <tinchodias@gmail.com> wrote:
On Sat, Sep 5, 2015 at 1:10 PM, Nicolai Hess <nicolaihess@web.de> wrote:
The same for the ]lang[ tag (see issue 15750)
Ok. But I still don't know how to add a url to a method's source code or comment from the UI.
I mean, people with a non-English keyboard can produce source code with ]lang[, but how can I produce a ]style[ ?
MartÃn
The whole mechanism is quite broken in Pharo4 upwards:
On Tue, Sep 8, 2015 at 12:29 PM, Henrik Johansen <henrik.s.johansen@veloxit.no> wrote:
On 08 Sep 2015, at 11:03 , Martin Dias <tinchodias@gmail.com> wrote:
On Sat, Sep 5, 2015 at 1:10 PM, Nicolai Hess <nicolaihess@web.de> wrote:
The same for the ]lang[ tag (see issue 15750)
Ok. But I still don't know how to add a url to a method's source code or comment from the UI.
I mean, people with a non-English keyboard can produce source code with ]lang[, but how can I produce a ]style[ ?
MartÃn
The whole mechanism is quite broken in Pharo4 upwards:
Not sure what to do in Pharo: parse the Text or just ignore the ]style[, but at least file-in should not raise an error. I reported it: https://pharo.fogbugz.com/f/cases/16575/CodeImporter-doesn-t-parse-style-chu... MartÃn
Hi Martin, Sent from my iPhone
On Sep 5, 2015, at 3:06 AM, Martin Dias <tinchodias@gmail.com> wrote:
Hi all
On Sat, Sep 5, 2015 at 3:44 AM, Ben Coman <btc@openinworld.com> wrote:
On Sat, Sep 5, 2015 at 4:43 AM, Eliot Miranda <eliot.miranda@gmail.com> wrote:
On Sep 4, 2015, at 1:33 PM, stepharo <stepharo@free.fr> wrote:
Nobody defines urls in method comment anymore.
I do. Several places in the Cog JIT contain irks to specification documents. I don't care how its implemented (providing it's preserved in text, or maybe have the pretty-printer scan comments for http://. But it is sad that we can't have hyperlinks in text.
Look I know the implementation stinks, but it can be fixed. But if the functionality is lost the system is weakened. Instead of asking what you can get rid if why not ask what you can reimplement or enhance more elegantly.
Its a reasonable strategy to start the discussion at the extreme end and work back from there :) But if its being used we should think harder about it. What alternatives/workarounds are there?
First just to clarify, the aim is to remove Text from the image for the boot-strap?
Yes :) I heard that we want to do cut the dependency to Text-Core, so I browsed how to do it.
Ben, there is an explanation about ]style[ in the class comment of RemoteString.
Eliot, I didn't thought that somebody was using Text. In fact, I don't know how to do it in Pharo... how do you do?
I develop the VM in Squeak. There's a menu pick in the squeak text editor that lets you do it. In the hierarchy of emphasis types there's a class called something like TextURL. I'll take a look soon and check.
I found that ChangeRecord>>fileIn uses #newTextChunk to parse, but I just tested filing-in the attached .st file and it doesn't work. I guess the new CodeImporter replaced the ChangeRecord implementation, and it raises an error when ]style[.
MartÃn
1. Save the text internally as a string in the stream format and up-convert to Text only when its being displayed? But I guess these styles aren't restricted to just occurring in comments and maybe cause problem for compilation ??
2. As a stepping stone, introduce a TextPlaceholder class within minimal interface required by compilation which simply strips out all styles when Text is missing from the Image. When Text is present it #becomes back to it. Maybe risks loosing styles editing is done on the raw-stream-format. Maybe any editor would show the raw ]style[ and the compiler asks for #source which strips styles out.
3. Introduce Pillar and convert across. What are the requirements? a. URLs b. Class/method links c. Would you expect that Pillar can only appear in comments? Or in-line with code? d. Colouring? ...?
btw, I haven't come across this before. Can someone point me to some url & styled code where I can have a look at it
cheers -ben
Christ you're even using minis paced fonts. It's like the dark ages in there ;-)
In addition would prefer to have an explicit syntax as in Pillar or html to represent an hyperlink. We could have *Class>>#method* when pillar is not loaded and else a nicer representation. ctrl-click on a class should jump on it already and we could map it to *Class>>#method*
In class comments we should use pillar and get a real text renderer.
Hi Martin,
On Fri, Sep 4, 2015 at 8:13 AM, Martin Dias <tinchodias@gmail.com> wrote:
Hi all,
The source code of a method could be a Text, with TextAttributes. In such case, it will be written down to a stream with a "]style[" suffix. Through a special parsing, the system can reconstruct the original Text. I think the use of Text in source code or class comments is discouraged, but I understand we kept support for some backward compatibility. Now, may be it's time to remove this feature.
With Pablo, we looked for uses of ]style[ in the .sources and .changes files of a new Pharo 5.0, and there are no real uses. It appears in some comments only, as examples.
We tried to find past discussions about this feature, because we're sure there were, but we couldn't.
So, the questions is:
Can we remove support for text in source code or class comments in Pharo 5.0?
IIUC, ]style[ also supports things like embedding URL links and method references in class comments. So I would request that it /not/ be removed. It is very useful to put clickable links in class comments.
I think I can propose a slice.
Regards, MartÃn PS: I was trying to cut dependencies of Text-Core with System-Changes and System-Changes.
-- _,,,^..^,,,_ best, Eliot
<HelpOnHelp class-example.st>
2015-09-04 21:27 GMT+02:00 Eliot Miranda <eliot.miranda@gmail.com>:
Hi Martin,
On Fri, Sep 4, 2015 at 8:13 AM, Martin Dias <tinchodias@gmail.com> wrote:
Hi all,
The source code of a method could be a Text, with TextAttributes. In such case, it will be written down to a stream with a "]style[" suffix. Through a special parsing, the system can reconstruct the original Text. I think the use of Text in source code or class comments is discouraged, but I understand we kept support for some backward compatibility. Now, may be it's time to remove this feature.
With Pablo, we looked for uses of ]style[ in the .sources and .changes files of a new Pharo 5.0, and there are no real uses. It appears in some comments only, as examples.
We tried to find past discussions about this feature, because we're sure there were, but we couldn't.
So, the questions is:
Can we remove support for text in source code or class comments in Pharo 5.0?
IIUC, ]style[ also supports things like embedding URL links and method references in class comments. So I would request that it /not/ be removed. It is very useful to put clickable links in class comments.
I missed that too, why did we remove that?
I think I can propose a slice.
Regards, MartÃn PS: I was trying to cut dependencies of Text-Core with System-Changes and System-Changes.
-- _,,,^..^,,,_ best, Eliot
I think that having metadata (style) language and data (source code) mixed together is a bad idea. I never like the ]lang[ tag because it is a huge hack. It does not even exist in the Smalltalk syntax!!! We save code that the parser cannot parse. What we fun idea. So people are bashing for backward compatibility and we remove a bad way to encode metadata then suddenly it looks like we were doing something bad. Stef
Hi Stef,
On Sep 21, 2015, at 2:53 PM, stepharo <stepharo@free.fr> wrote:
I think that having metadata (style) language and data (source code) mixed together is a bad idea. I never like the ]lang[ tag because it is a huge hack. It does not even exist in the Smalltalk syntax!!! We save code that the parser cannot parse. What we fun idea. So people are bashing for backward compatibility and we remove a bad way to encode metadata then suddenly it looks like we were doing something bad.
It seems to me you are confusing functionality and implementation. Just because you don't like the implementation is no reason to lose the functionality. No implementation is perfect, so this path leads to a denuded system. The onus is on the community to reimplement things well, not throw things away just because the implementation is problematic. If I had chosen to implement Cog with a much nicer substrate then neither Squeak, nor Pharo nor Newspeak would have been able to benefit. Instead I was pragmatic and made sure Cig supported the existing requirements, not done platonic ideal. I wish you would adopt as a principle that functionality be retained as you are cleaning up Pharo. Instead I see useful fictional its discarded because if scruples and I find it "unpleasant". _,,,^..^,,,_ (phone)
Stef
(autocorrect-unmangled version... ;) Eliot writes:
...I was pragmatic and made sure Cog supported the existing requirements, not some platonic ideal. I wish you would adopt as a principle that functionality be retained as you are cleaning up Pharo. Instead I see useful functionality is discarded because of scruples and I find it "unpleasant".
-C -- Craig Latta netjam.org +31 6 2757 7177 (SMS ok) + 1 415 287 3547 (no SMS)
On 21 Sep 2015, at 23:53, stepharo <stepharo@free.fr> wrote:
I think that having metadata (style) language and data (source code) mixed together is a bad idea. I never like the ]lang[ tag because it is a huge hack. It does not even exist in the Smalltalk syntax!!! We save code that the parser cannot parse. What we fun idea. So people are bashing for backward compatibility and we remove a bad way to encode metadata then suddenly it looks like we were doing something bad.
Stef
I am with Stef, it is a silly idea to mix the two. Nobody uses this in Pharo. Cleaning up means simplifying too.
Le 22 sept. 2015 07:40, "Sven Van Caekenberghe" <sven@stfx.eu> a écrit :
On 21 Sep 2015, at 23:53, stepharo <stepharo@free.fr> wrote:
I think that having metadata (style) language and data (source code)
mixed together is a bad idea.
I never like the ]lang[ tag because it is a huge hack. It does not even exist in the Smalltalk syntax!!! We save code that the parser cannot parse. What we fun idea. So people are bashing for backward compatibility and we remove a bad way to encode metadata then suddenly it looks like we were doing something bad.
Stef
I am with Stef, it is a silly idea to mix the two. Nobody uses this in Pharo. Cleaning up means simplifying too.
In the code no. But in the comments, that would be good to have back. In color form. As we can write pillar class comments, can't we render them ? Moose as an editor/viewer for pillar files. Phil
Le 22 sept. 2015 à 07:49, philippe.back@highoctane.be a écrit :
Le 22 sept. 2015 07:40, "Sven Van Caekenberghe" <sven@stfx.eu> a écrit :
On 21 Sep 2015, at 23:53, stepharo <stepharo@free.fr> wrote:
I think that having metadata (style) language and data (source code) mixed together is a bad idea. I never like the ]lang[ tag because it is a huge hack. It does not even exist in the Smalltalk syntax!!! We save code that the parser cannot parse. What we fun idea. So people are bashing for backward compatibility and we remove a bad way to encode metadata then suddenly it looks like we were doing something bad.
Stef
I am with Stef, it is a silly idea to mix the two. Nobody uses this in Pharo. Cleaning up means simplifying too.
In the code no. But in the comments, that would be good to have back. In color form. As we can write pillar class comments, can't we render them ? Moose as an editor/viewer for pillar files
Yes, it is where we want to go: use Pillar for styled comments. Once we have that, we can remove Text styling from source code / comments.
On 22 Sep 2015, at 08:47, Christophe Demarey <Christophe.Demarey@inria.fr> wrote:
Le 22 sept. 2015 à 07:49, philippe.back@highoctane.be a écrit :
Le 22 sept. 2015 07:40, "Sven Van Caekenberghe" <sven@stfx.eu> a écrit :
On 21 Sep 2015, at 23:53, stepharo <stepharo@free.fr> wrote:
I think that having metadata (style) language and data (source code) mixed together is a bad idea. I never like the ]lang[ tag because it is a huge hack. It does not even exist in the Smalltalk syntax!!! We save code that the parser cannot parse. What we fun idea. So people are bashing for backward compatibility and we remove a bad way to encode metadata then suddenly it looks like we were doing something bad.
Stef
I am with Stef, it is a silly idea to mix the two. Nobody uses this in Pharo. Cleaning up means simplifying too.
In the code no. But in the comments, that would be good to have back. In color form. As we can write pillar class comments, can't we render them ? Moose as an editor/viewer for pillar files
Yes, it is where we want to go: use Pillar for styled comments. Once we have that, we can remove Text styling from source code / comments.
Yes, inside a comment, which should be totally opaque from a language syntax standpoint, you can do whatever you want: HTML, Markdown, Pillar, Pier, ...
2015-09-22 9:10 GMT+02:00 Sven Van Caekenberghe <sven@stfx.eu>:
On 22 Sep 2015, at 08:47, Christophe Demarey < Christophe.Demarey@inria.fr> wrote:
Le 22 sept. 2015 à 07:49, philippe.back@highoctane.be a écrit :
Le 22 sept. 2015 07:40, "Sven Van Caekenberghe" <sven@stfx.eu> a écrit
:
On 21 Sep 2015, at 23:53, stepharo <stepharo@free.fr> wrote:
I think that having metadata (style) language and data (source
code) mixed together is a bad idea.
I never like the ]lang[ tag because it is a huge hack. It does not even exist in the Smalltalk syntax!!! We save code that the parser cannot parse. What we fun idea. So people are bashing for backward compatibility and we remove a bad way to encode metadata then suddenly it looks like we were doing something bad.
Stef
I am with Stef, it is a silly idea to mix the two. Nobody uses this in Pharo. Cleaning up means simplifying too.
In the code no. But in the comments, that would be good to have back. In color form. As we can write pillar class comments, can't we render them ? Moose as an editor/viewer for pillar files
Yes, it is where we want to go: use Pillar for styled comments. Once we have that, we can remove Text styling from source code / comments.
Yes, inside a comment, which should be totally opaque from a language syntax standpoint, you can do whatever you want: HTML, Markdown, Pillar, Pier, ...
But, if you want to integrate with IDE support for that extended syntax, then it would be nice to have an idea of the syntax used... Otherwise we're back at the special leading char to determine the format :( Side opinion to come to the point: I think that Pillar would make a better implementation than the ]style[ stuff. Thierry
Hi, I think that having Pillar in comments is definitely the way to go. In: http://www.humane-assessment.com/blog/writing-pillar-books-with-the-gtinspec... I show also an extension that renders comments with PIllar highlighting. This is working now in an inspector, but the same principle can be applied to other browsers. Cheers, Doru On Tue, Sep 22, 2015 at 8:47 AM, Christophe Demarey < Christophe.Demarey@inria.fr> wrote:
Le 22 sept. 2015 à 07:49, philippe.back@highoctane.be a écrit :
Le 22 sept. 2015 07:40, "Sven Van Caekenberghe" <sven@stfx.eu> a écrit :
On 21 Sep 2015, at 23:53, stepharo <stepharo@free.fr> wrote:
I think that having metadata (style) language and data (source code)
mixed together is a bad idea.
I never like the ]lang[ tag because it is a huge hack. It does not even exist in the Smalltalk syntax!!! We save code that the parser cannot parse. What we fun idea. So people are bashing for backward compatibility and we remove a bad way to encode metadata then suddenly it looks like we were doing something bad.
Stef
I am with Stef, it is a silly idea to mix the two. Nobody uses this in Pharo. Cleaning up means simplifying too.
In the code no. But in the comments, that would be good to have back. In color form. As we can write pillar class comments, can't we render them ? Moose as an editor/viewer for pillar files
Yes, it is where we want to go: use Pillar for styled comments. Once we have that, we can remove Text styling from source code / comments.
-- www.tudorgirba.com "Every thing has its own flow"
On Tue, Sep 22, 2015 at 9:13 AM, Tudor Girba <tudor@tudorgirba.com> wrote:
Hi,
I think that having Pillar in comments is definitely the way to go.
In:
http://www.humane-assessment.com/blog/writing-pillar-books-with-the-gtinspec...
I show also an extension that renders comments with PIllar highlighting. This is working now in an inspector, but the same principle can be applied to other browsers.
Yes, that's what I tried in the past. The ability to associate graphics explaining structures is really useful there. Phil
Cheers, Doru
On Tue, Sep 22, 2015 at 8:47 AM, Christophe Demarey < Christophe.Demarey@inria.fr> wrote:
Le 22 sept. 2015 à 07:49, philippe.back@highoctane.be a écrit :
Le 22 sept. 2015 07:40, "Sven Van Caekenberghe" <sven@stfx.eu> a écrit :
On 21 Sep 2015, at 23:53, stepharo <stepharo@free.fr> wrote:
I think that having metadata (style) language and data (source code)
mixed together is a bad idea.
I never like the ]lang[ tag because it is a huge hack. It does not even exist in the Smalltalk syntax!!! We save code that the parser cannot parse. What we fun idea. So people are bashing for backward compatibility and we remove a bad way to encode metadata then suddenly it looks like we were doing something bad.
Stef
I am with Stef, it is a silly idea to mix the two. Nobody uses this in Pharo. Cleaning up means simplifying too.
In the code no. But in the comments, that would be good to have back. In color form. As we can write pillar class comments, can't we render them ? Moose as an editor/viewer for pillar files
Yes, it is where we want to go: use Pillar for styled comments. Once we have that, we can remove Text styling from source code / comments.
-- www.tudorgirba.com
"Every thing has its own flow"
participants (13)
-
Ben Coman -
Christophe Demarey -
Craig Latta -
Eliot Miranda -
Henrik Johansen -
Martin Dias -
Nicolai Hess -
phil＠highoctane.be -
philippe.back＠highoctane.be -
stepharo -
Sven Van Caekenberghe -
Thierry Goubier -
Tudor Girba