Committed changes to my bpython fork. Likely, these changes can be committed to upstream, but its probably better to test them. I suspect that it may be best not to include oranj support for the main bpython, but instead do the following:
Move the python.py file to a languages package within the bpython folder Add a command line switch -l which takes an argument, a language name, and imports languages.xxx as language. Then, one could install additional languages by dumping their files into that directory. This would fix several problems associated with the current oranj support (it requires oranj itself, which is currently hard to install) and also means not distributing other modules such as OranjLexer. I'll spend a bit more time looking for changes to make (I think that there are a few more things to language-agnostify), and then hopefully prepare a nice patch / ask for commit privilages. On May 13, 8:18 pm, Bob Farrell <[email protected]> wrote: > So, after a long time trying to get everything running (largely due to being > distracted on IRC, granted), I managed to track down and fix the problems. > > Before I start, I'd like to say I'm really impressed with the work you've done > - I nearly had a shit fit when I saw you'd gutted all of the attribute lookup > stuff and wondered why on earth you were surprised it was not working, and > then > noticed the "python" module and dependency injection stuff you're doing. So, > good work ! > > Now, here's why you weren't finding the problem: > try: > self.completer.complete(cw, 0) > except Exception: > # This sucks, but it's either that or list all the exceptions that could > # possibly be raised here, so if anyone wants to do that, feel free to send me > # a patch. XXX: Make sure you raise here if you're debugging the completion > # stuff ! > > Unfortunately, given that all kinds of exceptions can occur during completion > stuff (since we're letting the user execute code inside our program), a long > time ago I resigned to just catching "Exception" and dealing with it - > TypeError, ValueError, blah blah blah, can all be raised. So, unless you do a > "raise" in that except block, anything that goes wrong in the completion stuff > will be ignored and you won't see any legitimate bugs in the code. > > In this case, firstly the "types" module hadn't been imported in python.py, > secondly "matches" doesn't get defined here if attr_lookup fails (this is > technically a bug in my code): > try: > matches = attr_lookup(obj, expr, attr) > finally: > # Dark magic: > if __getattribute__ is not None: > setattr(type_, '__getattribute__', __getattribute__) > if __getattr__ is not None: > setattr(type_, '__getattr__', __getattr__) > # /Dark magic > return matches > > so that raises a NameError, and thridly, "expr" and "attr" weren't being > passed > in to get_obj_attrs, so they were undefined too. attr_matches now passes them > in: > > def attr_matches(self, text): > """Taken from rlcompleter.py and bent to my will.""" > > m = re.match(r"(\w+(\.\w+)*)\.(\w*)", text) > if not m: > return [] > > expr, attr = m.group(1, 3) > obj = eval(expr, self.interp.locals) > > s = language.get_obj_attrs(obj, expr, attr) > return s > > (Previously just did "language.get_obj_attrs(obj)"). > > Okay ! So, I've attached a patch; everything seems to work fine now; hopefully > you'll be able to use what I've done to get it working with oranj - just make > sure to raise that exception I mentioned ! > > Oh, and that "foo(" stuff in the completion list is pretty cool; looking > forward to pulling that upstream. > > As far as I can tell everything's ready to go as-is as far as pulling changes > upstream goes. Once we've got that done I'm happy to give you commit privs to > the main bpython repo so you can push your changes there directly - you've > done > a great job on the language-agnostic front. Can you think of anything that's > preventing me pulling upstream now ? I'll let you get back to me before I do > anything. > > On Wed, May 13, 2009 at 12:01:12PM -0700, Pavel Panchekha wrote: > > > Committed new revision to oranj. Try to redownload it and modify the > > file src/objects/about.py to the installation path (or just read the > > INSTALL file). > > > I'll get a build script up and running for that at some point. > > > On May 13, 5:19 pm, Bob Farrell <[email protected]> > > wrote: > > > Been fighting for a while now to get this working but there are a some > > > missing > > > dependencies. > > > > I'm guessing this: > > > http://code.activestate.com/recipes/475116/ > > > is the "terminal" module. > > > > No idea what the "files" module is, but it's looking for that too, and I > > > really > > > need to get out of the office now. I might take a look at this later but > > > ideally if you could get this working in a --no-site-packages virtualenv > > > and > > > list any dependencies on the bitbucket.org wiki (or include any extra > > > deps in > > > the distribution) that'd be great. > > > > Nice work, though - good to see you're taking the right approach with > > > keeping > > > bpython agnostic. Once we figure out what's up with the attribute lookup > > > issue > > > (which I can probably diagnose by just using bpython without the --oranj > > > flag) > > > I'll do some testing and then I see no problem pulling your changes > > > upstream. > > > > On Wed, May 13, 2009 at 08:20:32AM -0700, Pavel Panchekha wrote: > > > > >http://bitbucket.org/pavpanchekha/bpython-oranj/ > > > > > Python is now a bit broken (no attr matching), oranj just as good (use > > > > --oranj command line flag). > > > > > Because I've yet to make a build system thing for oranj, the following > > > > directory tree is required: > > > > > oranj > > > > - folder > > > > - bpython > > > > - src > > > > - ... > > > > > folder can be called anything. > > > > > I should really fix that. > > > > > On May 13, 2:22 pm, Bob Farrell <[email protected]> > > > > wrote: > > > > > Well, there's some hideously dark magic that goes on with the > > > > > autocompletion > > > > > stuff which you might want to bypass. Feel free to send a patch over > > > > > (or, > > > > > ideally, if you can fork the bpython project on bitbucket.org so I > > > > > can pull and > > > > > push without having to worry about emailing patch files around) and > > > > > I'll take a > > > > > look. I'm at work, may as well do something interesting. :) > > > > > > On Wed, May 13, 2009 at 05:58:47AM -0700, Pavel Panchekha wrote: > > > > > > > Hmm. I have it mostly working, but I seem to have completely broken > > > > > > attribute matching. Don't know why, and can't seem to be able to > > > > > > debug > > > > > > anything. If you want, I can send you a patch as it is now, maybe > > > > > > you > > > > > > will understand better than me. Don't want to waste your time > > > > > > though. > > > > > > > What I have right now is that I split out all of the language > > > > > > specific > > > > > > stuff completely, and everything but the attribute matching works > > > > > > (Oh, > > > > > > and I haven't tested rewind, but that shouldn't be a problem). > > > > > > > On May 13, 9:46 am, Bob Farrell > > > > > > <[email protected]> > > > > > > wrote: > > > > > > > Hi Pavel, > > > > > > > > On Tue, May 12, 2009 at 05:58:13PM -0700, Pavel Panchekha wrote: > > > > > > > > > Having a bit of trouble here with the color formatting. The > > > > > > > > BPythonFormatter works (mostly, had to fix it up a bit) but > > > > > > > > there's > > > > > > > > another issue. > > > > > > > > > When Oranj prints error messages, it'd be nice if they were > > > > > > > > colored. > > > > > > > > They already are when using the console, but I need to use the > > > > > > > > curses > > > > > > > > color codes for printing in the bpython. > > > > > > > > > What I'm doing is this: > > > > > > > > > msg = "\x01r\x03%s\x01d\x03%s" % (type(e).__name__, ("" if not > > > > > > > > e.args > > > > > > > > else (": " + " ".join(map(str, e.args))))) > > > > > > > > > Ideally, this should make the first bit (type(e).__name__) red > > > > > > > > and the > > > > > > > > other bit normal gray colored (Its the error message). This > > > > > > > > isn't > > > > > > > > working, however - the entire line is red. > > > > > > > > I thought this was already possible, but turns out it wasn't - > > > > > > > I've changed the > > > > > > > code so that you can achieve what you were aiming for now by > > > > > > > doing this: > > > > > > > msg = "\x01r\x03what\x04\x01d\x03hello" > > > > > > > and then passing that to Repl.write, which will now split on \x04 > > > > > > > - you'll have > > > > > > > to pull from the hg repo to get the latest changes, or just use > > > > > > > this: > > > > > > > > def write(self, s): > > > > > > > """For overriding stdout defaults""" > > > > > > > if '\x04' in s: > > > > > > > for block in s.split('\x04'): > > > > > > > self.write(block) > > > > > > > return > > > > > > > if s.rstrip() and '\x03' in s: > > > > > > > t = s.split('\x03')[1] > > > > > > > else: > > > > > > > t = s > > > > > > > > if isinstance(t, unicode): > > > > > > > t = t.encode(getattr(sys.__stdout__, 'encoding') or > > > > > > > sys.getdefaultencoding()) > > > > > > > > if not self.stdout_hist: > > > > > > > self.stdout_hist = t > > > > > > > else: > > > > > > > self.stdout_hist += t > > > > > > > > self.echo(s) > > > > > > > self.s_hist.append(s.rstrip()) > > > > > > > > The more I look at the way the colours work in bpython, the more > > > > > > > I hate it. Oh > > > > > > > well. :) > > > > > > > > > Anything I'm doing wrong? I could send complete, (sorta) > > > > > > > > working code > > > > > > > > if you want, but in any case, is there anything obviously wrong? > > > > > > > > > On May 12, 5:06 pm, Pavel Panchekha <[email protected]> > > > > > > > > wrote: > > > > > > > > > Thanks for the rundown. It shouldn't be too hard - most of > > > > > > > > > the stuff > > > > > > > > > you mentioned is already there in a half-done way, just need > > > > > > > > > to clean > > > > > > > > > it up. With luck, I'll extract the python-specific bits as I > > > > > > > > > go, so it > > > > > > > > > can be merged upstream. I'll try to keep the mailing list > > > > > > > > > posted if I > > > > > > > > > finish up. > > > > > > > > > > On May 12, 4:53 pm, Bob Farrell > > > > > > > > > <[email protected]> > > > > > > > > > wrote: > > > > > > > > > > > Hi Pavel, > > > > > > > > > > > On Tue, May 12, 2009 at 11:01:05AM -0700, Pavel Panchekha > > > > > > > > > > wrote: > > > > > > > > > > > > I'd like to write a shell for oranj similar to bpython. > > > > > > > > > > > I've already > > > > > > > > > > > written a pygments parser for it, and writing completion > > > > > > > > > > > shouldn't be > > > > > > > > > > > too hard. What else needs to be changed before I can get > > > > > > > > > > > a full > > > > > > > > > > > bpython-esque shell for oranj? Is there anything in > > > > > > > > > > > curses to fiddle > > > > > > > > > > > with, for example? (I tried > > ... > > read more » > > attrlookup.patch > 2KViewDownload --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "bpython" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/bpython?hl=en -~----------~----~----~----~------~----~------~--~---
