[vw7.7] Internationalization bug in NumberReader>>readLatinNumberFrom:type?

Hi,

 

I'm trying to use

               

                Locale current currencyPolicy reader readFrom: aStream type: FixedPoint

 

to read in a currency value from a stream. There is a problem with it because the method eventually calls NumberReader>>readLatinNumberFrom:type: and this one gives special meaning to the characters 'deqs' it may encounter during parsing. These characters are given the meaning of a 'coercionClass'. 'e' means Floating point exponent character; 'd' and 'q' mean Double exponent character and 's' means FixedPoint exponent character.

 

Does this interpretation belong in this method?  I would think this is more a feature wanted in programming language parsing, but not to be used in Locale-dependent parsing methods?

 

I want to use it to be able for an end user to enter a monetary value followed by a currency symbol, for instance:

 

                10000 euro

 

Parsing this works fine as long as there is a space between 10000 and 'euro'. But if the user enters '10000euro' without a space then the NumberReader first extracts '10000e' as a floating point number and then the rest of the stream contains only 'uro', which leads to incorrect results.

 

Do you think this is a bug?

 

I'm thinking of overriding this method and removing the coercing part. Or is there another way to achieve what I want to do?

 

Thanks,

Mark

 

NumberReader>>readLatinNumberFrom: aStream type: typeClass

 

                | negative whole precision fractional eChar possibleCoercionClass exp coercionClass saved |

                negative := (aStream peekFor: $-)

                                                               ifTrue: [-1]

                                                               ifFalse: [1].

                precision := nil -> nil.

                coercionClass := Integer.

                whole := self getIntegerPartFrom: aStream digits: precision.

                saved := precision value.

                (aStream peekFor: printPolicy decimalPoint)

                               ifTrue:

                                               [coercionClass := Float.

                                               fractional := self getIntegerPartFrom: aStream digits: precision.

                                               (precision value = 0 and: [saved = 0])

                                                               ifTrue: [self error: (#errNoNumber << #dialogs >> 'No number found.')]]

                               ifFalse:

                                               [precision value = 0

                                                               ifTrue: [self error: (#errNoNumber << #dialogs >> 'No number found.')].

                                               fractional := 0.

                                               precision value: 0].

                eChar := aStream peek.

                eChar == nil

                               ifTrue: [possibleCoercionClass := nil]

                               ifFalse:

                                               [possibleCoercionClass := self chooseFloatRepresentationFor: eChar.

                                               possibleCoercionClass == nil ifFalse: [aStream next]].

                exp := nil.

                possibleCoercionClass == nil

                               ifFalse:

                                               [| endOfNumber neg |

                                               coercionClass := possibleCoercionClass.

                                               endOfNumber := aStream position.

                                               neg := aStream peekFor: $-.

                                               (printPolicy isDigit: aStream peek)

                                                               ifTrue:

                                                                              [exp := self getIntegerPartFrom: aStream digits: nil asValue.

                                                                              neg ifTrue: [exp := exp negated]]

                                                               ifFalse: [aStream position: endOfNumber]].

                ^typeClass

                               coerce: fractional / (10 ** precision value) + whole * negative

                               to: coercionClass

                               precision: precision value

                               exponent: exp

                               exponentChar: eChar

Thanks Les,

 

I'll make an override then and I can reevaluate when the NumberReader has been reviewed.

 

If you're going to reconsider NumberReader, perhaps one thing you could do is to provide a different way for error handling compared to what happens now? In the current implementation, when something goes wrong an Error exception is thrown:

 

                self error: (#errNoNumber << #dialogs >> 'No number found.')

 

But this is a very general error exception. I think it would be better to throw a different kind of exception so that the handler can be sure that it's a parsing error that occurred instead of some programming error. This exception could contain extra information regarding the position where something went wrong for instance.

 

Best regards,

Mark