On 02 Jun 2014, at 02:06, José Comesaña <jose.comes...@gmail.com> wrote:
> One last thing. If you do: > > sd := STON fromString: '1345.76s2'. > sd2 := ScaledDecimal readFromString: '1345.76s2'. > > you can note that sd = sd2 returns false, and that sd - sd2 returns 0.00s2. > Both numbers are different in the order of precision, because the first has > numerator 5918715072783319 and denominator 4398046511104 and the second has > numerator 33644 and denominator 25. I still have to fix this. The problem was that #parseNumberFraction returned a Float which let to reduced precision when creating the ScaledDecimal. Now it returns a pure/precise Fraction, with Float conversion happening later, if necessary. > Hope not to bore you too much. No, you do not ;-) > Regards > > > > 2014-06-02 1:56 GMT+02:00 José Comesaña <jose.comes...@gmail.com>: > And this is an improved version. More changes to avoid parsing weird things > as 123.456s3e2 (it could be improved, for sure, but I don't have any more > time...) > > parseNumber > | negated number | > negated := readStream peekFor: $-. > number := self parseNumberInteger. > (readStream peekFor: $.) > ifTrue: [ number := number + self parseNumberFraction ]. > (readStream peekFor: $s) > ifTrue: [ | scale | > scale := self parseNumberInteger. > number := ScaledDecimal newFromNumber: number scale: > scale ] > ifFalse: [ > (readStream peekFor: $/) > ifTrue: [ | denominator | > denominator := self parseNumberInteger. > number := Fraction numerator: number > denominator: denominator ] > ifFalse: [ > ((readStream peekFor: $e) or: [ > readStream peekFor: $E ]) > ifTrue: [ number := number * > self parseNumberExponent ]. > ] > ]. > negated > ifTrue: [ number := number negated ]. > self consumeWhitespace. > ^ number > > > 2014-06-02 1:51 GMT+02:00 José Comesaña <jose.comes...@gmail.com>: > > Hi Sven. > > First of all, I know that in VisualWorks there exists a class called > FixedPoint, that is (more or less) similar to ScaledDecimal (is what the > class comment says). But, instead of 123.45s2, VWs writes 123.45s. > NEVERTHELESS, it can read 123.45s2 with no pain. > > There exists a Fraction type also in VW, which, I believe, uses the same > notation as in Pharo. > > I know that portability is an important issue, but, at least for me, > portability from Pharo to Pharo is the most important. I depend on > ScaledDecimal for an application (for storing money values -changing the > storage mode is not an option now-) and need the possibility of converting it > to string and reading it back. The string version is my database, that gets > loaded into memory upon program start. > > You can see here the changes I have made to STONReader >>#parseNumber: > > parseNumber > | negated number | > negated := readStream peekFor: $-. > number := self parseNumberInteger. > (readStream peekFor: $.) > ifTrue: [ number := number + self parseNumberFraction ]. > " -------------- New from here -------------- " > (readStream peekFor: $s) > ifTrue: [ | scale | > scale := self parseNumberInteger. > number := ScaledDecimal newFromNumber: number scale: > scale ]. > (readStream peekFor: $/) > ifTrue: [ | denominator | > denominator := self parseNumberInteger. > number := Fraction numerator: number denominator: > denominator ]. > " -------------- to here -------------- " > ((readStream peekFor: $e) or: [ readStream peekFor: $E ]) > ifTrue: [ number := number * self parseNumberExponent ]. > negated > ifTrue: [ number := number negated ]. > self consumeWhitespace. > ^ number > > I am a newbie about STON. I don't know if it is correct, but it works for > me, at least for some tests I have made. I know it needs improvements, > thinking about strange strings we could receive. But I trust the writer not > to create such weird things. > > I could suggest another way of making the function of parseNumber (OK, OK, I > know I am thinking from a Pharo-only perspective...). Seeing that number > parsing always ends with this condition: > > readStream atEnd not and: [ readStream peek isDigit ] > > why not create a temporary string (including possibility of $e, $s, $/, even > $@ ...) to that point (tmpString) and do: > > number := Number readFromString: tmpString > > That way, we would have a "native" parsing in each language > > Hope this helps. If someone needs more information, pleas ask. > > Best regards > > > > 2014-05-31 21:29 GMT+02:00 Sven Van Caekenberghe <s...@stfx.eu>: > > Hi José, > > On 31 May 2014, at 01:32, José Comesaña <jose.comes...@gmail.com> wrote: > > > Wouldn't it be good if STON could save ScaledDecimals as 12345.678s?. That > > way, we could read them back again as ScaledDecimal, instead of Float. I > > have tried and it seems quite useful. > > > > Regards > > That is an interesting idea and it is probably nice to have for those relying > on ScaledDecimals. > > One of the ideas of STON is to be portable across dialects, so I am wondering > if ScaledDecimals exists everywhere ? > > I am curious as to how you did it though, since STONReader basically contains > its own number parser. Could you share your code ? > > Sven > > >