Yes, one day we have to do this. There is also Xtreams which is a broader, more interesting way of rethinking the subject.
But the main issue is: the stream API is way too broad, so any change will involve slimming down or changing the API and hence the users. That is a lot of work. > On 18 Dec 2014, at 23:51, stepharo <[email protected]> wrote: > > Sven > > could we not nuke > > MultiByteFileStream > > and use ZnCharacterStreams? > > > Le 17/12/14 19:54, Sven Van Caekenberghe a écrit : >>> On 17 Dec 2014, at 19:42, [email protected] wrote: >>> >>> Hi there, >>> >>> I am still working my way through MultiByteFileStream tests, and have come >>> across the messages #converter and #converter: . I would like to be able >>> to test these - could someone explain briefly how the converters are used >>> in this class? >> In order to convert from bytes to characters and vice versa, >> MultiByteFileStream uses a TextConverter subclass instance as a helper. The >> messages #nextFromStream: and #nextPut:onStream: are the main interface >> between the two. >> >> The silly thing is that both MultiByteFileStream and TextConverter convert >> from wrongly encoded strings to correctly encoded strings. >> ZnCharacter[Read|Write]Stream and ZnCharacterEncoder do the correct thing. >> >> Sven >> > >
