site_archiver@lists.apple.com Delivered-To: darwin-dev@lists.apple.com Ah, but the structure and keys of the output are not documented, and AFAIK have changed somewhat over the ages. -- Terry _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev@lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.appl... On May 15, 2008, at 2:10 PM, Finlay Dobbie wrote: On Thu, May 15, 2008 at 10:08 PM, Jordan K. Hubbard <jkh@apple.com> wrote: On May 15, 2008, at 1:51 PM, Finlay Dobbie wrote: Is the format of output from commandline utilities guaranteed across software updates? I'm sure jkh has made statements to the effect that they are not :o) Unless a command-line utility is producing XML or some other format specifically designed and documented to be "machine readable", then the format is essentially whatever we say it is from release to release and there are no warrantees expressed or implied. Finlay is correct and Terry was over-simplifying things just a bit in that regard. lsof and ps are documented as producing machine parseable output, if the appropriate command line arguments are used; for lsof, this is "- F"; for ps, this is "-o". "system_profiler -xml" produces XML output; "system_profiler" without - xml is documented as producing colon separated key/value pairs, when queried within a single subhierarchy, which was the specific case under discussion (i.e. also machine parseable for the restricted case being discussed). Obviously, if a command doesn't have the necessary options, you are at more risk parsing its output -- but less risk than if you were to (for example) use sysctl-based data interfaces or grovel around in /dev/mem or /dev/kmem, with your code depending on kernel symbols and data structures not changing. They will change. You're right, that one defect of XML is the lack of tool-specific DTDs. You can parse output into an associative array, but you can't guarantee me that a key/value pair isn't going to go away from revision to revision, since there is no DTD to nail me to a cross of immutability. Thus the only thing you can really guarantee me is that the parsing code can be normalized and thus simplified. So, Finlay, what do you suggest as a more perfect solution to the problem? This email sent to site_archiver@lists.apple.com