2011/1/7 Henrik Rudstrøm <hrudst...@googlemail.com> > > On Jan 7, 2011, at 3:24 PM, Thomas Paviot wrote: > > 2011/1/7 Henrik Rudstrom <hrudst...@googlemail.com> > >> Hello all, >> > > Hi Henrik, > > >> I've been watching this thread from the sideline for a while, very >> interesting thread, but i havent had time to share my thoughts yet. so here >> we go. >> Although the idea of a standardized api across platforms and all the >> possibilities for collaboration/online services has a lot of potential, i >> think it will become a very big undertaking and i agree with Dave that we >> should limit our scope and start simple. >> >> Although im not against it, i see some dangers with building a new api on >> top of open cascade, even if we start with a very delimited scope; it will >> be a lot of work to develop our own set of sematics (the OMG CAD looked >> interesting though, but i got the impression that opencascade has much more >> functionality than what is covered by OMG) >> It will be difficult to cover everything in a consistent manner, so the >> risk is having still the granular c++ like opencascade api with some apis >> built on top that are good for some purposes but not all, this can make it >> very daunting for new users (and still for me) to figure out when i can use >> a nice abstracted function or when i have to go down to 'pure' opencascade. >> > > I agree with all the points you mentioned. You're especially right dealing > with the "granularity" issue. The big deal is how to go from low level data > models (very granular, like the STEP data model, or even OpenCascade) to > high level functions (or services), less granular, without loosing any > semantics. > > Mostly when the high level api subject comes up on the list it seems that >> most of the issues people have are with the programming paradigms of >> OpenCascade, not necessary the semantics of opencascade. In my opinion, if >> we could strip away all c++ artifacts of opencascade i think the semantics >> of opencascade as it is is quite decent. >> > > I understand. If the OCC semantics is good enough for your needs, let's say > it's ok. There might be a confusion I'd like to talk to you about our: the > lack of a clean, simple, explicit and shared CAD API is a major issue that > still has to be solved. This is a 'research' point of view (which is part of > my daily job). On the other side, we all agree that the OpenCascade API is > huge, with weird naming conventions, and that an overall clean-up is > required. It's not practical. This is a 'pythonocc developer' viewpoint. > Regarding the first point (the issue) and the second point (pythonocc is not > practical, poorly documented etc.), I'd like to join these efforts. > > > It is indeed a bit confusing, i think the term High Level API leaves a lot > of room for interpretation; high level in terms of CAD operations or in > terms of programming. it might be a bit too general term. >
I would name 'high level in terms of operations' as 'CAD Services', but the name already exists (OMG CAD Services). These services must be computer independent and can be implemented as webservices or a python API (or another language API, scripting engine etc). The python API implementing the services would be the 'high level API' built on top of pythonocc. Also if High Level refers to High level CAD operations im also a bit curious > what you (and others) see as (generic) high level operations? correct me if > im wrong, but once you get beyond the basic functionality of a solid > modelling kernel things start to become more specific to certain types of > design/production methods etc. > Absolutely. The cad services/high level api only focuses on 'basic functionalities' as you say, or generic CAD operations independent from any specific business case (whether it is automotive industry, space industry, architecture etc). Indeed, 'high level' is a relative concept, not absolute. It would be an intermediate level between a low level API (i.e. OCC, STEP) and business specific operations (eg make_car(wheels=4), make_plane(wings=2, wing_length=23m), make_building(floors=4, windows_per_floor=21) etc. these are really stupid examples!). > If High level referes to high level in terms of programming paradigms, ie > abstract away all the boilerplate code/cleanup the api (a loss-less > compression of the api if you will) would give a good overview for further > iteration. > I love this analogy with lossles compression. Your proposal of refactoring the wrapper would lead to a small compression ratio. The services/high level api compression ratio would be higher. > > > However, I don't want to impose my views here. This ml is a > 'pythonocc-users' ml, so if pythonocc users just request for a polish of the > API, let's go with it. > > . > I just speak for myself ;) > > maybe i can clarify my point by first quoting you earlier in this thread: > "Regarding the second point (basic primitives such as point, line, circle > etc. or other entities like coord_sys, group etc.), this is exactly what > we're thinking about for the 'High Level API' (HLA) for pythonOCC. In my > opinion, it is the most important part of the pythonOCC project, since we > all agree that the OCC library is too granular in order to be easily, > quickly and efficiently deployed." > My main point is that opencascade already has these entities, but they are > kind of buried in boilerplate code and wierd syntax. I'm just a bit afraid > that it might become messy if we start building a paralell structure of > entities on top of the existing one. > > > Maybe the recent discussions related to high level api, as I understand it, > or the db storage, have nothing to do in this ml. I would understand if I > bore anyone with this kind of stuff. A pythonocc-r&d or pythonocc-dev ml may > be better. If only one, two or three people here are interested in > discussing a midterm or longterm future of the project, let's move away. We > could then leave this ml to practical discussions about how tos, faqs etc. > and small API improvements. > > Dont get me wrong, i dont find it a bore at all quite the opposite, but > sometimes it is not so clear what the purpose of the discussion actually > is. > > > But one thing is sure : I will work on that topic for the first 6 months of > 2011 in my research activity. Maybe I have to wait for a few months before > coming back here with solutions and proposals, rather than questions. > > > :) > > > >> >> A while back i was playing around with pygccxml (a c++ parser that can be >> used for code generation) to get an idea of what could be done to make >> opencascade more user friendly without having to manually write all the >> wrappers. There are quite a lot of things that could be done with the >> wrappers to make them more compact and 'pythonic' without manually writing >> all the code: >> >> - Handles (i believe Thomas managed already to get rid of those, havent >> tested that) >> - The collections classes: Array1Of.... the methods of these classes could >> be rewritten to conform to the list/set/etc interfaces of python and all >> methods that accept these collections could have some aout generated code >> that converts standard python lists to these classes if necessary (i already >> have the code to identify all the collection classes) >> - The strictness of arguments in general could be loosened or some classes >> could be merged ( for example gp_Pnt/gp_XYZ/gp_Vec/gp_Dir) are basically the >> same thing >> - Geom and gp could be merged or autoconverted between >> - A lot of the algorithms classes that need to be initialized, executed >> and then a method call to return the object of interest follow similar >> patterns that could merged into a single method call, for example the >> sequence: >> cut = BRepAlgoAPI_Cut(shape1, shape2) >> cut.Build() >> cutted_shape=cut.Shape() >> - There still are a lot of method calls that takes a pointer to an object >> that serves as the return value >> - We could add a simple regexp to rewrite the CapitalizedMethodNames to >> pythonic_underscored_method_names >> - Eliminate the prefix of all the classes and implement them as python >> namespaces: >> Geom_Point -> Geom.Point or >> from Geom import * >> p = Point(...) >> - then there is a lot of adapters etc that i havent looked too much at. >> >> Many of these patterns can be identified with pygccxml queries and used to >> add code to the wrappers. Some would require manual specification of classes >> and methods that belong to a pattern. >> >> Im not sure i would call it a high level api anymore, but i think it would >> alleviate a lot of the need for a general high level api. >> >> what do you guys think? >> > > I love your pygccxml/SWIG suggestions. Get ready to commit! > > > I think i sent you some of that code sometime after the pythonocc > 'conference' in oslo this last year (which was based on your existing swig > generator code but i was so rude to refactor it beyond recognition) > I can remember receiving your code, but I have to retrieve it from my messy folder hierarchy (hope google mail will help me)! > Im getting in the mood to revive it, but i'd like to know that if it will > fit in the big picture and that my assumptions are right before i spend too > much time on it. > > > > >> >> Henrik >> > > Thomas > > > Henrik > Thomas
_______________________________________________ Pythonocc-users mailing list Pythonocc-users@gna.org https://mail.gna.org/listinfo/pythonocc-users
