On Fri, Sep 9, 2011 at 11:35, Thomas Goirand <[email protected]> wrote: > Hi, > > Few months ago, I had a try with Google pinyin. But I gave up, because > it wasn't good enough: it crashed, and couldn't guess words based on the > one you typed before, because upstream code simply didn't have the > feature that wasn't ported from the Java version. At the time, > GooglePinyin wasn't remotely comparable to the Windows version. >
This Google Pinyin is the one works on android devices, it's not designed to compare with the one on Windows, but I should admit this one is better than ibus-pinyin, and sunpinyin prior to 2.0.4 (not released yet). It was unstable on Linux, and I guess it was because the program is designed for a different libc - android does not use glibc/eglibc which we use. Now libgooglepinyin has fixed most things so it's quite stable now. > So, I am wondering, is GooglePinyin better now, and can it be remotely > compared to SunPinyin, which is well maintained already, and that I've > been using for a very long time? Aren't we seeing too many input methods > in Debian, for no real benefits (just see how many pinyin methods we > have...)? > When it becomes much more stable, you'll find it's far better than Sun Pinyin (especially before 2.0.4, which has been tried integrated things from googlepinyin). I'm not sure about the details of input model, but AFAIK the dict of Sunpinyin is very poor comparing with googlepinyin's one. > I'm not willing to point fingers at here, just trying to understand what > you are doing, making sure that GooglePinyin is valuable enough, and > that we don't have better options already (eg: sunpinyin, which already > has a pretty good grammar analysis to predict words). > > Thomas > In fact we don't have a better dict than googlepinyin's one in open source world, and we don't have another lightweight solution. libgooglepinyin is very small in size (the binary is no more than 400KiB, the dict is only 1.5MiB, while others have several MiB to tens of MiB), and has better user experience than any other existing input methods, so I believe it's worthwhile. -- Regards, Aron Xu -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
