----- "Michael DeHaan" <mdehaan@xxxxxxxxxx> wrote: > Adrian Likins wrote: > > > > Between Red Hat Summit, FUDCon, and various discussions, it's > occurred > > to me that lots of people > > are using the 'func' command line in applications instead of the > > python api. This is particular common > > in non python apps of course, where the api isn't useable. > > > > The problem with the current commandline client is that it is kind > of > > limited usefulness in this > > useage. Parsing output seems to be an issue. Current versions > support > > the use of the "--xmlrpc", > > "--json", and "--pickle" output modes, which could theoretically be > > > used to make this a little easier. > > It wouldn't be hard to add other ways to marshall more complex data > > > either, yaml, for example. > > Or perhaps a mode that would be easily parsed with commandline tools > > > (csv? BAR=foo shellvar style?) > > Any ideas? > > > > The other side to the problem is the input. At the moment, the > "call" > > support only supports passing string > > args on the command line. There are a couple of ways to make that > more > > flexible that come to mind. One > > is to add some support for specifing types. Not sure what that would > > > look like at the moment. > > > > Another idea that I like, is the ability to read marshalled data in > > > from stdin. It would be pretty easy to > > support xmlrpc style data this way, or json, or other types. It > might > > make it a little easier to glue support > > for other languages in this way, and not be constrained by what can > be > > passed into the command line > > +1 > > > > A wrapper api could use native support for xmlrpc/json/yaml/etc to > > build up the data structures, and then > > feed then via stdin to func. > > Luca/Symbolic guys, what do you think of the above? I think it's > easier than the XMLRPC over socket route and perhaps a bit easier to put > together on for a language that doesn't have very good access to OS > level bits (read: java). Func still requires permissions on the > certs, so xmlrpc marshalling over STDIN seems fair to me, and other > apps are known to do it. > > The other easy thing to do would be YAML, but we lose the ability to > express faults if we go that route. it seems a good feature because simplify in a good way the managing of return data from functions, however I think call back-end through launch a command is a bit dirty, but is definetely the quickest and simplest way, so +1 :) I think to continue also my develop on a client/server implementation (I hope to finish before the end of the world) so we have another way to communicate to func. it's ok or is useless? bye Luca _______________________________________________ Func-list mailing list Func-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/func-list
