ERft - return false (not throw an exception) if function hasArray( .. ) is used with an
non-existing rftreport.
ERst - new memeber function hasArray ( .. )
Convert the Python opm package from a regular package to a namespace
package such that opm-simulators can contribute to the package from a
different filesystem path. In this way, the two packages opm.simulators
and opm.io (for example) can have a different parent filesystem path.
Previously, touching CMakeLists.txt after a first compilation would
trigger compile errors as the two CMake runs were done with
inconsistent parameters (first one without USE_SUPERLU defined and
second one with -DUSE_SUPERLU=ON). Of course that meant that the
options would also differ from module to module, which seems like a
recipe for desaster.
This patch fixes this by moving the definition of the USE_SUPERLU
option to OpmInit.cmake which is included very early in the top most
CMakeLists.txt of each module.
ClosesOPM/opm-simulators#3908
We need to honour a "small rate threshold" when outputting the lift
gas max supply rate limit and total gas production limit. This
threshold renders the output discontinuous, but is needed for
compatibility.
Uses GNUInstallDIrs' CMAKE_INSTALL_DATAROOTDIR as top directory for
common python scripts. Also adds that location to
opm-project-config.cmake.in such that other opm modules can
easily lookup the common scripts
Loading summary vector from restart simulation with loadBaseRunData = true and selected summary vector not present in baserun.
Common use case where a new vector added in the restart run.
The ESMRY file can be updated with more time step data in the time span between opening an ESMRY file
and loading actual summary data from the file. This can happen if the ESMRY is generated by an active
run. Length of summary vectors extracted will be equal to number of time steps on disk when the file was
first opened (with the constructor).
We effectively ignore JFUNC data unless end-point scaling is
activated through the ENDSCALE keyword in the RUNSPEC section.
Alert the user to this fact by terminating the simulation run if the
case uses JFUNC without ENDSCALE present in RUNSPEC. In this case,
issue a diagnostic message of the form
Error: Unrecoverable errors while loading input: Problem with keyword JFUNC
In CASE.DATA line 125
Incompatible keyword combination: JFUNC declared, but ENDSCALE is missing.
The current Network::Balance class applies default values a little
too eagerly and this leads to problems for restarted simulation. We
must take care to preserve the "did this item have a value" state
also in the restart file lest interactions with TUNING be lost
there.
To this end, make the minimum timestep size for network balancing be
an std::optional<> and use std::nullopt to represent "no value set"
in the input file. While here, also preserve negative balancing
intervals exactly instead of replacing these values by zero.
In particular, add more unit testing and give names to a number of
index constants.
Thanks to [at]blattms for reviewing the initial version of this
patch.
Instead, just declare that all non-FIELD groups are subject to
higher-level control of the water injection.
This hack is needed for compatibility on a real field case. We must
revisit this at a later point.
If a simulation model explicitly assigns a target or limit of 0.0,
then we must honour that value in the restart file. Use a more
complete condition to detect whether or not group level targets or
limits are defaulted in GCONPROD and/or GCONINJE.
In particular, add entries from WECON, WGRUPCON and, WVFPEXP and fix
an incorrect item attribution for the WTEST 'reason' parameter.
Thanks to [at]tskille for invaluable assistance in identify these
items.
