The Trunk: Multilingual-ar.87.mcz
Locked
 
|
Andreas Raab uploaded a new version of Multilingual to project The Trunk:
http://source.squeak.org/trunk/Multilingual-ar.87.mcz ==================== Summary ==================== Name: Multilingual-ar.87 Author: ar Time: 11 February 2010, 11:48:33.736 pm UUID: ca61be1a-7e30-2840-9b77-9cc7ecfc2d3f Ancestors: Multilingual-ar.86 Change EncodedCharSet>>digitValue: to EncodedCharSet>>digitValueOf:. Part 3/3. =============== Diff against Multilingual-ar.86 =============== Item was changed: ----- Method: EncodedCharSet class>>digitValueOf: (in category 'class methods') ----- digitValueOf: char "Answer 0-9 if the receiver is $0-$9, 10-35 if it is $A-$Z, and < 0 otherwise. This is used to parse literal numbers of radix 2-36." | value | value := char charCode. + value <= $9 asciiValue ifTrue: + [^value - $0 asciiValue]. + value >= $A asciiValue ifTrue: + [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]. + (value >= $a asciiValue and: [value <= $z asciiValue]) ifTrue: + [^value - $a asciiValue + 10]]. - value <= $9 asciiValue - ifTrue: [^value - $0 asciiValue]. - value >= $A asciiValue - ifTrue: [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]]. ^ -1 ! Item was changed: ----- Method: Unicode class>>digitValueOf: (in category 'class methods') ----- digitValueOf: char "Answer 0-9 if the receiver is $0-$9, 10-35 if it is $A-$Z, and < 0 otherwise. This is used to parse literal numbers of radix 2-36." | value | value := char charCode. + value <= $9 asciiValue ifTrue: + [^value - $0 asciiValue]. + value >= $A asciiValue ifTrue: + [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]. + (value >= $a asciiValue and: [value <= $z asciiValue]) ifTrue: + [^value - $a asciiValue + 10]]. - value <= $9 asciiValue - ifTrue: [^value - $0 asciiValue]. - value >= $A asciiValue - ifTrue: [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]]. value > (DecimalProperty size - 1) ifTrue: [^ -1]. ^ (DecimalProperty at: value+1) ! Item was removed: - ----- Method: LanguageEnvironment class>>digitValue: (in category 'accessing') ----- - digitValue: char - - ^ Unicode digitValue: char. - ! Item was removed: - ----- Method: EncodedCharSet class>>digitValue: (in category 'class methods') ----- - digitValue: char - "Answer 0-9 if the receiver is $0-$9, 10-35 if it is $A-$Z, and < 0 - otherwise. This is used to parse literal numbers of radix 2-36." - - | value | - value := char charCode. - value <= $9 asciiValue - ifTrue: [^value - $0 asciiValue]. - value >= $A asciiValue - ifTrue: [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]]. - ^ -1 - ! Item was removed: - ----- Method: Unicode class>>digitValue: (in category 'class methods') ----- - digitValue: char - - | value | - value := char charCode. - value <= $9 asciiValue - ifTrue: [^value - $0 asciiValue]. - value >= $A asciiValue - ifTrue: [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]]. - - value > (DecimalProperty size - 1) ifTrue: [^ -1]. - ^ (DecimalProperty at: value+1) - ! |
Re: The Trunk: Multilingual-ar.87.mcz
|
|
Of course, as indicated by Bert, this make some test fail.
If we follow this road, then we must also remove support for reading/printing Float in non decimal base. I personnally agree that this form is mostly educational and has no major industrial utility. As squeak is also an educational support, I will let other debate if this is a major feature or not... I'm not at all in favour in decomposing operations as you suggested earlier because this introduces additional round off errors: | a b | a := SqNumberParser parse: '3r1.1e55'. b := (SqNumberParser parse: '3r1.1') * (3 raisedTo: 55). a - b -> 3.4359738368e10 If you use current squeak (Number>>#readFrom:) then above forms are equivalent. This is because (Number>>#readFrom:) use the exact broken decomposition you suggested. This decomposition is broken because it does not answer the nearest Floating point value to the literal expression. Proof: 3r1.1e55 is literally same as 3r11e54, that is (3+1)*(3 raisedTo: 54). Of course, this number cannot be represented exactly as a Float, it has too many digits: | a | a := (3+1)*(3 raisedTo: 54). a highBit - a lowBit -> 85, that means a 85 bit mantissa is required (plus implied leading 1), Float only has 52 bits (plus implied leading 1) SqNumberParser does answer a Float nearest the exact literal than Number>>readFrom: ((SqNumberParser parse: '3r1.1e55') asTrueFraction - ((3r11)*(3 raisedTo: 54))) abs < ((Number readFrom: '3r1.1e55') asTrueFraction - ((3r11)*(3 raisedTo: 54))) abs -> true In fact, the nearest Float, as you can check with predecessor successor. | a exact | a := SqNumberParser parse: '3r1.1e55'. exact := (3r11)*(3 raisedTo: 54). self assert: (a - exact) abs <= (a successor - exact) abs. self assert: (a - exact) abs <= (a predecessor - exact) abs. This of course fail with Number>>readFrom: You can also simply try 3r1.1e55 - 3r11.0e54 and the SqNumberParser form... (SqNumberParser parse: '3r1.1e55') - (SqNumberParser parse: '3r11.0e54') Maybe it's time to push the readFrom: changes i already pushed in Pharo... Nicolas 2010/2/12 <[hidden email]>: > Andreas Raab uploaded a new version of Multilingual to project The Trunk: > http://source.squeak.org/trunk/Multilingual-ar.87.mcz > > ==================== Summary ==================== > > Name: Multilingual-ar.87 > Author: ar > Time: 11 February 2010, 11:48:33.736 pm > UUID: ca61be1a-7e30-2840-9b77-9cc7ecfc2d3f > Ancestors: Multilingual-ar.86 > > Change EncodedCharSet>>digitValue: to EncodedCharSet>>digitValueOf:. Part 3/3. > > =============== Diff against Multilingual-ar.86 =============== > > Item was changed: > ----- Method: EncodedCharSet class>>digitValueOf: (in category 'class methods') ----- > digitValueOf: char > "Answer 0-9 if the receiver is $0-$9, 10-35 if it is $A-$Z, and < 0 > otherwise. This is used to parse literal numbers of radix 2-36." > > | value | > value := char charCode. > + value <= $9 asciiValue ifTrue: > + [^value - $0 asciiValue]. > + value >= $A asciiValue ifTrue: > + [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]. > + (value >= $a asciiValue and: [value <= $z asciiValue]) ifTrue: > + [^value - $a asciiValue + 10]]. > - value <= $9 asciiValue > - ifTrue: [^value - $0 asciiValue]. > - value >= $A asciiValue > - ifTrue: [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]]. > ^ -1 > ! > > Item was changed: > ----- Method: Unicode class>>digitValueOf: (in category 'class methods') ----- > digitValueOf: char > "Answer 0-9 if the receiver is $0-$9, 10-35 if it is $A-$Z, and < 0 > otherwise. This is used to parse literal numbers of radix 2-36." > > | value | > value := char charCode. > + value <= $9 asciiValue ifTrue: > + [^value - $0 asciiValue]. > + value >= $A asciiValue ifTrue: > + [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]. > + (value >= $a asciiValue and: [value <= $z asciiValue]) ifTrue: > + [^value - $a asciiValue + 10]]. > - value <= $9 asciiValue > - ifTrue: [^value - $0 asciiValue]. > - value >= $A asciiValue > - ifTrue: [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]]. > > value > (DecimalProperty size - 1) ifTrue: [^ -1]. > ^ (DecimalProperty at: value+1) > ! > > Item was removed: > - ----- Method: LanguageEnvironment class>>digitValue: (in category 'accessing') ----- > - digitValue: char > - > - ^ Unicode digitValue: char. > - ! > > Item was removed: > - ----- Method: EncodedCharSet class>>digitValue: (in category 'class methods') ----- > - digitValue: char > - "Answer 0-9 if the receiver is $0-$9, 10-35 if it is $A-$Z, and < 0 > - otherwise. This is used to parse literal numbers of radix 2-36." > - > - | value | > - value := char charCode. > - value <= $9 asciiValue > - ifTrue: [^value - $0 asciiValue]. > - value >= $A asciiValue > - ifTrue: [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]]. > - ^ -1 > - ! > > Item was removed: > - ----- Method: Unicode class>>digitValue: (in category 'class methods') ----- > - digitValue: char > - > - | value | > - value := char charCode. > - value <= $9 asciiValue > - ifTrue: [^value - $0 asciiValue]. > - value >= $A asciiValue > - ifTrue: [value <= $Z asciiValue ifTrue: [^value - $A asciiValue + 10]]. > - > - value > (DecimalProperty size - 1) ifTrue: [^ -1]. > - ^ (DecimalProperty at: value+1) > - ! > > > |
Re: The Trunk: Multilingual-ar.87.mcz
|
|
Nicolas Cellier wrote:
> This of course fail with Number>>readFrom: > > You can also simply try > 3r1.1e55 - 3r11.0e54 > and the SqNumberParser form... > (SqNumberParser parse: '3r1.1e55') - (SqNumberParser parse: '3r11.0e54') > > Maybe it's time to push the readFrom: changes i already pushed in Pharo... Go for it. You're the best expert we have in this area, and I trust your judgment. Cheers, - Andreas |
|
|
Also see
http://code.google.com/p/pharo/issues/detail?id=1258 '.1' asNumber toss an error in Pharo 1.1 11043 and '1.0e+2' asNumber returns 1.0 yet other I think more knowledgeable people say http://www.wolframalpha.com/input/?i=1.0e%2B2 100 I'd just say any expectations of converting a string to a number in most flavours of Squeak should be viewed with suspicion. On 2010-02-12, at 12:44 AM, Andreas Raab wrote: > Nicolas Cellier wrote: >> This of course fail with Number>>readFrom: >> You can also simply try >> 3r1.1e55 - 3r11.0e54 >> and the SqNumberParser form... >> (SqNumberParser parse: '3r1.1e55') - (SqNumberParser parse: '3r11.0e54') >> Maybe it's time to push the readFrom: changes i already pushed in Pharo... > > Go for it. You're the best expert we have in this area, and I trust your judgment. > > Cheers, > - Andreas > -- =========================================================================== John M. McIntosh <[hidden email]> Twitter: squeaker68882 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com =========================================================================== |
Re: The Trunk: Multilingual-ar.87.mcz
|
|
2010/2/12 John M McIntosh <[hidden email]>:
> Also see > http://code.google.com/p/pharo/issues/detail?id=1258 > > '.1' asNumber toss an error in Pharo 1.1 11043 > and > > '1.0e+2' asNumber > returns 1.0 > If it's a matter of traditional Smalltalk syntax, we could check result with these Smalltalks: - gst - VW - Dolphin - vast - stx - ... If it's a matter of extending the syntax, we could also try current squeak version for: '+1' asNumber. '1,964' asNumber. Where to stop ? We have to write some EBNF rules and/or simply provide a few TestCase documenting the expectations (say with unit TestCase). In Pharo, I rejected undocumented non-Smalltalk syntax. The reason is that (Number>>readFrom:) is used by Parser, and 1.e2 or .1e2 are ambiguous in Smalltalk syntax. Stupid example: | e2 y | e2:=1. y:=1.e2 storeOn: Transcript.Transcript endEntry. ^y According to st-80 and documented Squeak EBNF syntax, it should print 1, not 100. This does not need to be written on stone, we can evolve Smalltalk syntax, or we can easily dissociate String>>asNumber from Parser usage with two subclasses of NumberParser... Some applications would also prefer a localized decimal separator for example... But this has to be decided on rational bases, not just by addition of incomplete undocumented hacks in readFrom: for specific application usage. Nicolas > yet other I think more knowledgeable people say > http://www.wolframalpha.com/input/?i=1.0e%2B2 > 100 > > I'd just say any expectations of converting a string to a number in most flavours of Squeak should be viewed with suspicion. > > On 2010-02-12, at 12:44 AM, Andreas Raab wrote: > >> Nicolas Cellier wrote: >>> This of course fail with Number>>readFrom: >>> You can also simply try >>> 3r1.1e55 - 3r11.0e54 >>> and the SqNumberParser form... >>> (SqNumberParser parse: '3r1.1e55') - (SqNumberParser parse: '3r11.0e54') >>> Maybe it's time to push the readFrom: changes i already pushed in Pharo... >> >> Go for it. You're the best expert we have in this area, and I trust your judgment. >> >> Cheers, >> - Andreas >> > > -- > =========================================================================== > John M. McIntosh <[hidden email]> Twitter: squeaker68882 > Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com > =========================================================================== > > > > > > |
Re: The Trunk: Multilingual-ar.87.mcz
|
|
In reply to this post by johnmci
On Friday 12 February 2010 02:40:36 pm John M McIntosh wrote:
> '1.0e+2' asNumber > returns 1.0 Squeak (Trunk) shows: '1.0e2' asNumber 100 '1.0e+2' asNumber 1.0 :-/ .. Subbu |
«
Return to Squeak - Dev
|
1 view|%1 views
| Free forum by Nabble | Edit this page |