> On 05 Jul 2016, at 16:40, Peter Uhnák <i.uh...@gmail.com> wrote: > > I know that only NeoCSV uses it — that's how I ran into this problem. I was > processing some (czech) CSV files which used the decimal comma separator… > however the numbers were silently truncated, which wasn't nice to say the > least — I really don't understand why the default behavior is to silently > change the value, and not produce an error — this also applies to Pharo's > number parser. > > BTW, you not only need to set the thousands separator, but the decimal > separator too, I guess. > > depending on the default values, but that's really not the main point > > Now, I can understand where/how your suggestions would make sense. Maybe you > can try subclassing and make your own variant (first) ? > > Well I would need a way to configure the CSV parser. Because I am certainly > not interested in manually transforming every float field. I want just > configure it at one place and use the regular addFloatField — after all the > file is going to be consistent in it's format. > > Btw there are other options for improvement, like configuring the default > date field and then having addDateField, etc. But maybe that's just > overloading the NeoCSV parser… in any case it's a food for thought.
Indeed, I do not want to overload the CSV parser, it is pretty simple right now. The conversions are all in the convenience protocol for a reason: they just save you some typing. You really ought to do your own conversions, when you need to. parser addFieldConverter: [ :string | MyNumberParser parse: string ] There are too many formats out there (especially for dates/times). You are right about truncation and error handling. But parsing and enforcing a syntax are two different things. That is why I think the thousands separator option is not that simple, consider 1,000.00 10,00.00 1,0,0,0.00 1,000.00E1000,000 You see ? One quick and dirty solution would be to just remove $, or replace one character by another. > Peter > > > On Tue, Jul 5, 2016 at 2:34 PM, Sven Van Caekenberghe <s...@stfx.eu> wrote: > Peter, > > NeoNumberParser is a simple number (integer/float) parser that is part of > NeoCSV (it was based on the JSON number parsing code). It was added because I > wanted a number parser that makes little demands on the stream it parses from > (just 1 character peek ahead, no arbitrary backtracking, limited API). It was > not meant to be very powerful. > > If you check the references, you see that where it is used in NeoCSVReader, > you could easily substitute another parser. > > Now, I can understand where/how your suggestions would make sense. Maybe you > can try subclassing and make your own variant (first) ? BTW, you not only > need to set the thousands separator, but the decimal separator too, I guess. > > Sven > > > On 05 Jul 2016, at 14:17, Peter Uhnák <i.uh...@gmail.com> wrote: > > > > Hi, > > > > is there any plan for NeoNumberParser do add localization support? > > > > e.g. > > > > NeoNumberParser new > > thousandsSeparator: $,; "common in us data" > > parse: '12,230' > > > > => > > > > 12230 > > > > NeoNumberParser new > > decimalSeparator: $,; "common in eu data" > > parse: '12,230' > > > > => > > > > 12.230 > > > > Thanks, > > Peter > > >
