Conditional based on whether or not a module is being used
in [Python]
|
Prev: Slicing [N::-1]
Next: ANN: Wing IDE 3.2.5 Released
From: Pete Emerson on 6 Mar 2010 18:13 On Mar 6, 2:38 pm, Vinay Sajip <vinay_sa...(a)yahoo.co.uk> wrote: > On Mar 5, 9:29 pm, Pete Emerson <pemer...(a)gmail.com> wrote: > > > > > I have written my first module called "logger" that logs to syslog via > > the syslog module but also allows forloggingto STDOUT in debug mode > > at multiple levels (to increase verbosity depending on one's need), or > > both. I've looked at theloggingmodule and while it might suit my > > needs, it's overkill for me right now (I'm still *very* much a python > > newbie). > > Overkill in what sense? You just need to write a few lines of code to > be able to use the logging package which comes with Python: > > import logging, logging.handlers, sys > logging.basicConfig(level=logging.DEBUG, stream=sys.stdout) > logging.getLogger().addHandler(logging.handlers.SysLogHandler()) > # default logs to syslog at (localhost, 514) with facility LOG_USER > # you can change the default to use e.g. Unix domain sockets and a > different facility > > So you're experienced enough and have time enough to write your own > logger module, but too much of a newbie to use a module which is part > of Python's included batteries? If you're writing something like > logging to learn about it and what the issues are, that's fair enough. > But I can't see what you mean by overkill, exactly. The three lines > above (or thereabouts) will, I believe, let you log to syslog and to > stdout...which is what you say you want to do. > > > I want to write other modules, and my thinking is that it makes sense > > for those modules to use the "logger" module to do thelogging, if and > > only if the parent using the other modules is also using the logger > > module. > > > In other words, I don't want to force someone to use the "logger" > > module just so they can use my other modules, even if the "logger" > > module is installed ... but I also want to take advantage of it if I'm > > using it. > > > Now that I've written that, I'm not sure that makes a whole lot of > > sense. It seems like I could say, "hey, this person has the 'logger' > > module available, let's use it!". > > > Thoughts? > > Well, the logging package is available in Python and ready for use and > pretty much battle tested, so why not use that? Are you planning to > use third-party libraries in your Python work, or write everything > yourself? If you are planning to use third party libraries, how would > their logging be hooked into your logger module? And if not, is it > good to have two logging systems in parallel? > > Of course as the maintainer of Python's logging package, you'd expect > me to be biased in favour of it. You maybe shouldn't let that sway > you ;-) > > Regards, > > Vinay Sajip Thanks for your insights, Vinay, and thank you also for writing packages such as logging. The word 'overkill' was a poor choice on my part! I should have said, "I don't quite understand the logging module yet, but I am comfortable with the syslog module's two functions, openlog and syslog". I wrote my own logger module *partly* to gain the experience, and partly to do the following: 1) In debug mode, send what would have gone to syslog to STDOUT or STDERR 2) In non-debug mode, use /dev/log or localhost:514 depending on what is set 3) Allow for multiple levels of logging beyond INFO, WARNING, CRIT ... essentially allow multiple levels of INFO depending on how much detail is desired. A high level of messaging when programs are running poorly is desired, but when programs are running smoothly, I don't need to send as much to syslog. I started in with your logging package, but I think I simply got ahead of myself. I definitely agree that writing my own wrappers around syslog to do what I want might be a duplication of effort. At this point I think I'm ready to go back to your logging package and see what I can do; if you have words of advice regarding 1-3 above, I'd certainly appreciate it. Now I'll go to your example above and see what it does. Thank you! Pete
From: Steve Holden on 7 Mar 2010 07:34 Vinay Sajip wrote: [...] > Well, the logging package is available in Python and ready for use and > pretty much battle tested, so why not use that? Are you planning to > use third-party libraries in your Python work, or write everything > yourself? If you are planning to use third party libraries, how would > their logging be hooked into your logger module? And if not, is it > good to have two logging systems in parallel? > > Of course as the maintainer of Python's logging package, you'd expect > me to be biased in favour of it. You maybe shouldn't let that sway > you ;-) > Vinay: My own impression of the logging module, formed from trying to use its documentation in the past, is that it's somewhat unapproachable, and difficult to use for simple purposes. I am happy to say that now I see the current (3.1) documentation it has improved to the point where I would be happy to try using it again. Thanks for your long-term maintenance of this package. regards Steve -- Steve Holden +1 571 484 6266 +1 800 494 3119 PyCon is coming! Atlanta, Feb 2010 http://us.pycon.org/ Holden Web LLC http://www.holdenweb.com/ UPCOMING EVENTS: http://holdenweb.eventbrite.com/
From: Jean-Michel Pichavant on 8 Mar 2010 08:02 Pete Emerson wrote: > On Mar 5, 1:14 pm, Chris Rebert <c...(a)rebertia.com> wrote: > >> On Fri, Mar 5, 2010 at 12:25 PM, Pete Emerson <pemer...(a)gmail.com> wrote: >> >>> On Fri, Mar 5, 2010 at 12:17 PM, Chris Rebert <c...(a)rebertia.com> wrote: >>> >>>> On 3/5/10, Pete Emerson <pemer...(a)gmail.com> wrote: >>>> >>>>> In a module, how do I create a conditional that will do something >>>>> based on whether or not another module has been loaded? >>>>> >> <snip> >> >>>>> If someone is using foo module, I want to take advantage of its >>>>> features and use it in foobar, otherwise, I want to do something else. >>>>> In other words, I don't want to create a dependency of foobar on foo. >>>>> >>>>> My failed search for solving this makes me wonder if I'm approaching >>>>> this all wrong. >>>>> >>>> Just try importing foo, and then catch the exception if it's not installed. >>>> >> <snip> >> >>> Except I want to use the module only if the main program is using it >>> too, not just if it's available for use. I think that I found a way in >>> my follow-up post to my own message, but not sure it's the best way or >>> conventional. >>> >> What is your use case for this behavior exactly? You've piqued my curiosity. >> >> Cheers, >> Chris >> --http://blog.rebertia.com >> > > I have written my first module called "logger" that logs to syslog via > the syslog module but also allows for logging to STDOUT in debug mode > at multiple levels (to increase verbosity depending on one's need), or > both. I've looked at the logging module and while it might suit my > needs, it's overkill for me right now (I'm still *very* much a python > newbie). > > I want to write other modules, and my thinking is that it makes sense > for those modules to use the "logger" module to do the logging, if and > only if the parent using the other modules is also using the logger > module. > > In other words, I don't want to force someone to use the "logger" > module just so they can use my other modules, even if the "logger" > module is installed ... but I also want to take advantage of it if I'm > using it. > > Now that I've written that, I'm not sure that makes a whole lot of > sense. It seems like I could say, "hey, this person has the 'logger' > module available, let's use it!". > > Thoughts? > The logging module is not overkill, use it. It can be simple if you want it to be, and can be more complex if you need advanced logging features. All the "what if the application is not using my logger module" is dealt with by the logging module. And I'm not its designer in any way, so my advice is completely objective :-) It's definitely worth spending some time ramping up with it. JM
From: Vinay Sajip on 9 Mar 2010 03:17 On Mar 7, 12:34 pm, Steve Holden <st...(a)holdenweb.com> wrote: > My own impression of theloggingmodule, formed from trying to use its > documentation in the past, is that it's somewhat unapproachable, and > difficult to use for simple purposes. > > I am happy to say that now I see the current (3.1) documentation it has > improved to the point where I would be happy to try using it again. > Thanks for your long-term maintenance of this package. > Hi Steve, Thanks for the positive feedback. The initial documentation for the logging package, because it lives in the library section of the overall documentation, was focused more on completeness of coverage for reference usage, rather than a more tutorial-based approach. Thanks to work by Doug Hellmann and others, the documentation has grown, over time, more accessible to Python novices. It's still not perfect, and I hope to be able to improve its clarity in the future, by getting help where possible from people who are better at technical writing than I am. I'm reviewing the documentation at the moment, as it happens, and it still seems hard to be able to put together a structure which is good for everyone. A full treatment, it seems to me, would talk a little about the detail of why things work as they do; but a lot of the time, people are just interested in getting going with the package, and less interested in the whys and wherefores. But for people trying to do more than the basics, that deeper understanding is sometimes necessary. The hard part is satisfying all audiences in one document! Regards, Vinay Sajip
From: Vinay Sajip on 9 Mar 2010 03:54
On Mar 6, 11:13 pm, Pete Emerson <pemer...(a)gmail.com> wrote: > > 1) In debug mode, send what would have gone to syslog to STDOUT or > STDERR > 2) In non-debug mode, use /dev/log or localhost:514 depending on what > is set > 3) Allow for multiple levels ofloggingbeyond INFO, WARNING, CRIT ... > essentially allow multiple levels of INFO depending on how much detail > is desired. A high level of messaging when programs are running > poorly is desired, but when programs are running smoothly, I don't > need to send as much to syslog. By "debug mode", do you mean the value of the __debug__ variable, or something else (e.g. a flag in your application)? You could certainly do something like (in your logging initialization code): if __debug__: handler = logging.StreamHandler() else: #use domain socket, UDP, etc. handler = logging.handlers.SocketHandler(...) logger.addHandler(handler) where logger is the root logger or some other high-level logger in your application. By the way, are you aware that accessing syslog via openlog etc. may not thread-safe, at least in some environments? Search the Web for "syslog openlog thread" for more info. You can certainly add additional levels to logging (see addLevelName), but I'm not sure that's what you really need. Generally, I find that when there are problems to be debugged, I get more benefits from using the logger hierarchy: I keep the level at logging.DEBUG but just log different things to different loggers. Just as a fr'instance, if I were logging the parsing of HTTP requests, I might use loggers named 'request', 'request.headers', 'request.headers.cookies', 'request.body', 'request.body.multipart' etc. When everything is working well, I have the verbosity of these loggers turned low by e.g. setting the level for the 'request' logger to WARNING or higher; when I want to debug header processing in more detail I might set the level of the 'request.headers' logger to DEBUG, which would output events from request header processing (but not the body), or just turn up the 'request.headers.cookies' level to look in more detail at what's happening during processing "Cookie:" headers. Regards, Vinay Sajip |