>the option is irrelevant for Resinsight.

Well - I will reiterate my case; we will burn that bridge when get there; last 
time I checked Resinsight was not using opm parser - and I am not aware of any 
plans to change that. Although I would be very pleased to be wrong on that 
point.

>I believe it is irrelevant for a number of simulator use cases too.

Well - then users of said simulator should take the offending keywords out of 
their deck.

>[...] so by design it makes more sense for the modules themselves to decide 
>whether to throw on unsupported options.

That way we distribute the definition of semantically correct out to the 
consumers - a consequence of that is that the "contract/promise" provided by 
the Parser becomes much weaker - I think that is unfortunate. An approach I 
would like/suggest was that the creation of EclipseState:: subobjects was more 
modular -  (on demand maybe??) - then problems would not arise before an 
unsupported property was actually used, that would be a bit like defaults are 
treated in the input deck: If a property without a default is defaulted - 
nothing will happen before you actually try to access that value.

But anyway - the GRIDFILE keyword is mostly interesting as a simple and current 
example; much more interesting are the properties which can be controlled both 
by sprinkling keywords through the SCHEDULE section, and by the 
opm-core/properties system. This includes the TUNING keywords and also the 
configuration of the restart output mentioned in PR821. I hope more people will 
read through this thread and voice an opinion.

J



From: Alf Birger Rustad
Sent: 3. juni 2015 15:17
To: Joakim Hove; [email protected]
Subject: RE: Improvements(?) in opm-parser

> We'll burn that bridge when we get there;

I believe we are already there, the option is irrelevant for Resinsight. I 
believe it is irrelevant for a number of simulator use cases too. My point 
being that keeping track of all modules and their use cases is already 
non-trivial, so by design it makes more sense for the modules themselves to 
decide whether to throw on unsupported options.

Cheers,
Alf

From: Opm [mailto:[email protected]] On Behalf Of Joakim Hove
Sent: 3. juni 2015 14:57
To: [email protected]<mailto:[email protected]>
Subject: Re: [OPM] Improvements(?) in opm-parser


>What if one downstream module supports it, but another doesn't, and for a 
>third module the option is irrelevant?
We'll burn that bridge when we get there; but quite frankly I do not see that 
we will ever get there. We (try) to adhere to the DRY principle in this part of 
town, and that means that capabilities like e.g. writing a grid is only 
implemented one single place - i.e. either Opm modules have access to a .GRID 
writer or they don't.  If application logic in one downstream module decides to 
not make use of that capability is of course another matter.

J
________________________________
From: Opm [[email protected]] on behalf of Joakim Hove 
[[email protected]]
Sent: Wednesday, June 03, 2015 12:39 PM
To: [email protected]<mailto:[email protected]>
Subject: Re: [OPM] Improvements(?) in opm-parser

There is of course an issue of priorities. Throwing will force us to prioritize 
a strict implementation of every keyword.
Well; let us walk through the GRIDFILE keyword - that is a good illustration of 
the situation:

1.       Opm simulators will output an .EGRID file unconditionally; that is 
also the default if no GRIDFILE keyword is present.

2.       In: https://github.com/OPM/opm-parser/pull/470 the parser is 
explicitly aware of the default behavior and the semantics of the GRIDFILE and 
NOGGRF keywords are correctly accounted.

3.       The old SPE decks explicitly ask for .GRID output - we do not support 
that.
At point 3 I see three possibilities:

1.       Throw an exception since the input deck asks for something we know the 
downstream modules can not support; to proceed the user should then update the 
input deck by modifying the GRIDFILE keyword - that has been done here: 
https://github.com/OPM/opm-data/pull/47

2.       Print a warning and don't give a shit.

3.       Implement support for .GRID output.
I strongly prefer alternative 1). I want the parser output to be taken 
seriously by the downstream modules; it should not be so that we configure 
EclipseState objects based on the input and then silently ignore them at will. 
If the common consensus is that for less important things like output 
configuration we do not really intend to follow the specifications in the input 
deck we should not even internalize that information - that might be a good 
alternative.
The parser should probably only be concerned with the parsing, and whether the 
parsed file is formatted correctly.
That is a long time ago; all of EclipseState is one semantic layer above this.

J


-------------------------------------------------------------------
The information contained in this message may be CONFIDENTIAL and is
intended for the addressee only. Any unauthorised use, dissemination of the
information or copying of this message is prohibited. If you are not the
addressee, please notify the sender immediately by return e-mail and delete
this message.
Thank you


-------------------------------------------------------------------
The information contained in this message may be CONFIDENTIAL and is
intended for the addressee only. Any unauthorised use, dissemination of the
information or copying of this message is prohibited. If you are not the
addressee, please notify the sender immediately by return e-mail and delete
this message.
Thank you


-------------------------------------------------------------------
The information contained in this message may be CONFIDENTIAL and is
intended for the addressee only. Any unauthorised use, dissemination of the
information or copying of this message is prohibited. If you are not the
addressee, please notify the sender immediately by return e-mail and delete
this message.
Thank you
_______________________________________________
Opm mailing list
[email protected]
http://www.opm-project.org/mailman/listinfo/opm

Reply via email to