I was looking for every valid character allowed as part of a binary selector name. Nothing in SelectorNode was obvious to me, but the AST diagram in my copy of Smalltalk-80, The Language and Implementation, says they are:
- + / * ~ < > = @ % | & ? ! Is this still correct? |
It also seems to indicate that - (minus) is only allowed all by itself, not with any of the others. The others are limited to a maximum of two. However, I was just able to save (compile) a method named -@+. So is it okay / advisable to go beyond the original spec if it lets me? On Sat, May 25, 2019 at 6:04 PM Chris Muller <[hidden email]> wrote:
|
Chris Muller wrote on Sat, 25 May 2019 18:08:28 -0500
> I was looking for every valid character allowed as part of a binary selector > name. Nothing in SelectorNode was obvious to me, but the AST diagram in > my copy of Smalltalk-80, The Language and Implementation, says they are: > - + / * ~ < > = @ % | & ? ! > Is this still correct? In code this is defined in method Scanner class>initialize where everything is a binary character by default, then tab lf ff cr space are redefined as delimiters, then the digits are redefined, then the letters and then " # $ ' : ( ) . ; [ ] { } ^ _ | each get their own token type. This means characters from 128 to 255 are also valid in binary selectors and even a bunch of control characters! > It also seems to indicate that - (minus) is only allowed all by itself, not with > any of the others. > The others are limited to a maximum of two. > However, I was just able to save (compile) a method named -@+. So is it > okay / advisable to go beyond the original spec if it lets me? This is defined in Scanner>>xBinary The two character limit is no longer there and the only restriction relative to $- is that it can't be immediately followed by a digit (in which case we can't tell if it is supposed to be the sign of the following number or the last character of the binary selector). I would avoid creating selectors that would not be valid in other Smalltalks if I might want to port my code later on. In the ANSI Smalltalk standard binary selectors can have any number of characters, but these must only be ! % & * + , / < = > ? @ \ ~ | - Binary selectors with the vertical bar are valid in ANSI but not in Squeak (they can mess things up with the block syntax and so aren't a good idea anyway). -- Jecel |
> On 27.05.2019, at 21:52, Jecel Assumpcao Jr. <[hidden email]> wrote: > > Chris Muller wrote on Sat, 25 May 2019 18:08:28 -0500 >> I was looking for every valid character allowed as part of a binary selector >> name. Nothing in SelectorNode was obvious to me, but the AST diagram in >> my copy of Smalltalk-80, The Language and Implementation, says they are: >> - + / * ~ < > = @ % | & ? ! >> Is this still correct? > > In code this is defined in method Scanner class>initialize where > everything is a binary character by default, then tab lf ff cr space are > redefined as delimiters, then the digits are redefined, then the letters > and then " # $ ' : ( ) . ; [ ] { } ^ _ | each get their > own token type. This means characters from 128 to 255 are also valid in > binary selectors and even a bunch of control characters! > >> It also seems to indicate that - (minus) is only allowed all by itself, not with >> any of the others. >> The others are limited to a maximum of two. >> However, I was just able to save (compile) a method named -@+. So is it >> okay / advisable to go beyond the original spec if it lets me? > > This is defined in Scanner>>xBinary > > The two character limit is no longer there and the only restriction > relative to $- is that it can't be immediately followed by a digit (in > which case we can't tell if it is supposed to be the sign of the > following number or the last character of the binary selector). > > I would avoid creating selectors that would not be valid in other > Smalltalks if I might want to port my code later on. In the ANSI > Smalltalk standard binary selectors can have any number of characters, > but these must only be ! % & * + , / < = > ? @ \ ~ | - > > Binary selectors with the vertical bar are valid in ANSI but not in > Squeak (they can mess things up with the block syntax and so aren't a > good idea anyway). Btw: × works just fine.... ¯\_(ツ)_/¯ Best regards -Tobias |
In reply to this post by Chris Muller-4
> So is it okay / advisable to go beyond the original spec if it lets me? I would prefer an interface that still looks like message sending in the first place. I suppose there are some cases where the parser (or compiler) let's you do something that might just be an oversight. I would consider "-@+" way too cryptic. That's the path you would be walking on if you follow that: self ><>. self »-(¯`·.·´¯)->. self --------{---(@. Not good. Just my two cents. Best, Marcel
|
In reply to this post by Jecel Assumpcao Jr
Thanks Jecel, that's exactly what I needed to know.
On Mon, May 27, 2019 at 2:52 PM Jecel Assumpcao Jr. <[hidden email]> wrote: > > Chris Muller wrote on Sat, 25 May 2019 18:08:28 -0500 > > I was looking for every valid character allowed as part of a binary selector > > name. Nothing in SelectorNode was obvious to me, but the AST diagram in > > my copy of Smalltalk-80, The Language and Implementation, says they are: > > - + / * ~ < > = @ % | & ? ! > > Is this still correct? > > In code this is defined in method Scanner class>initialize where > everything is a binary character by default, then tab lf ff cr space are > redefined as delimiters, then the digits are redefined, then the letters > and then " # $ ' : ( ) . ; [ ] { } ^ _ | each get their > own token type. This means characters from 128 to 255 are also valid in > binary selectors and even a bunch of control characters! > > > It also seems to indicate that - (minus) is only allowed all by itself, not with > > any of the others. > > The others are limited to a maximum of two. > > However, I was just able to save (compile) a method named -@+. So is it > > okay / advisable to go beyond the original spec if it lets me? > > This is defined in Scanner>>xBinary > > The two character limit is no longer there and the only restriction > relative to $- is that it can't be immediately followed by a digit (in > which case we can't tell if it is supposed to be the sign of the > following number or the last character of the binary selector). > > I would avoid creating selectors that would not be valid in other > Smalltalks if I might want to port my code later on. In the ANSI > Smalltalk standard binary selectors can have any number of characters, > but these must only be ! % & * + , / < = > ? @ \ ~ | - > > Binary selectors with the vertical bar are valid in ANSI but not in > Squeak (they can mess things up with the block syntax and so aren't a > good idea anyway). > > -- Jecel > |
In reply to this post by marcel.taeumel
> > So is it okay / advisable to go beyond the original spec if it lets me?
> > I would prefer an interface that still looks like message sending in the first place. I suppose there are some cases where the parser (or compiler) let's you do something that might just be an oversight. I would consider "-@+" way too cryptic. Yes, I wouldn't ever use it. It was just a test to see if the compiler would accept it. > That's the path you would be walking on if you follow that: > > self ><>. > self »-(¯`·.·´¯)->. > self --------{---(@. OMG, LOL!!!!!! > https://1lineart.kulaone.com/#/ :-) Wow. > > Not good. Just my two cents. Yes, definitely not good. I would never do it but my query is for how to take foreign code as an input, and I must convert binary selectors to an alphanumeric to conform to an underlying protocol, (while avoiding colliding with any keyword selectors), and with a reverse-conversion on the other end. So I needed to know all the possibilities to possibly expect for binary selectors, and even wondered whether I should disallow some based on some of those dimensions Jecel mentioned like standards. Cheers, Chris |
On 2019-05-29, at 7:54 PM, Chris Muller <[hidden email]> wrote:So is it okay / advisable to go beyond the original spec if it lets me? Of course, the burning question is whether unicode emoji char points are valid for unary/binary selectors? foo 👈 42 🧛♂️ + myThing 🇬🇧 thatThing ie foo := 42 vampire + myThing union thatThing 🧐 Almost Smalltalk-72 ? tim -- tim Rowledge; [hidden email]; http://www.rowledge.org/tim C++ is history repeated as tragedy. Java is history repeated as farce. |
Free forum by Nabble | Edit this page |