* Completions: Apply perforation cell filter to COMPDAT and MSW export
* PerforationInterval: Limit Cell Filter dropdown to case-level filters
* PerforationInterval: Prefix Cell Filter entries with case name when multiple cases
* Completions: Use std::format for cell filter suppression log messages
Adds a new HUD label in the upper-left of the 3D view that reads
"Filter : <name(s)>" whenever a cell, property, or per-view data
filter is active. The label is hidden when nothing is enabled.
Names are joined per the source collection's AND/OR combine mode, and
contributions from different collections are concatenated with ", ".
On Eclipse views the unified RimFilterInViewCollection is used as the
single anchor; GeoMech views aggregate cell and GeoMech property
filters directly.
Guard zero-length keywords in ecl_kw_max_min, increase grdecl buffer to 256 and bound sscanf widths, and validate dimensions and keyword sizes (COORD, ZCORN, MAPAXES, CORSNUM, ACTNUM) when loading GRDECL grids.
When the user toggles "Show Curve" off, updateCurvePresentation runs
updateCurveVisibility, which detaches and deletes m_plotCurve, before
calling updateCurveAppearance. The override in RimWellFlowRateCurve
dereferenced m_plotCurve unconditionally and crashed. Match the null
guard already present in RimPlotCurve::updateCurveAppearance.
Fixes#14011.
Nested Qt event processing during toolbar clearing can trigger
performUpdate while the visible-index-to-position mapping is out of
sync with the current visible plot widgets. The lookup lambda in
alignAxis previously asserted on a miss, calling abort() which the
SIGABRT handler reported as a crash. Return std::optional and skip
widgets without a mapping entry instead, matching the defensive
null-pointer style used elsewhere in this file.
Fixes#14008.
The CVF_ASSERT on the wellPathGeometry pointer aborted release builds
before the existing null check could return gracefully. A well path
without resolved geometry is a valid state, so drop the assert and let
the early return handle it.
Fixes#14010.
Three fixes from PR review on the data-filters refactor.
RimDataFilterInViewCollection::syncWithSource: avoid double-free on
project load when m_sourceCollection has not yet resolved. The first
pass deletes orphan wrappers but leaves them in the local snapshot;
the null-source branch then deleted them again. Replace the second
deletion with deleteChildren() on the wrapper field.
RimDataFilterInViewCollection::setSourceCollection: disconnect from
the previous source's filtersChanged signal before reassigning, so
edits to a previously-bound case no longer trigger spurious geometry
regen on this view.
RimEclipsePropertyFilter::fieldChangedByUi: always emit
triggerFilterChanged() after updating bounds and related fields.
Otherwise, when the property filter lives inside a combined filter
under a per-view RimEclipsePropertyFilterCollection, the combined
parent's auto-derived display name was never recomputed.
Adds case.filtered_cells(filter, time_step, grid_index) which applies any
RimCellFilter to an Eclipse case and returns a per-cell 0/1 mask. The mask
length and ordering match case.grid_property(..., grid_index), so callers
can combine the two vectors element-wise.
Implemented as a scriptable Rimc method (PdmVoidObjectMethod) on
RimEclipseCase that writes the mask via the existing key-value store, with
a thin Python wrapper mirroring the visible_cells() pattern. No proto or
gRPC service changes.
Lets a script attach a cell filter to a RimPerforationInterval as a
typed reference so subsequent work can gate completion-export
connections by the filter's cell mask. Pure metadata in this commit:
no consumer of RimPerforationInterval is modified, the field exists
purely as an API surface.
Add a non-scriptable caf::PdmPtrField<RimCellFilter*> m_cellFilter to
RimPerforationInterval with public cellFilter()/setCellFilter()
accessors and a "Cell Filter" group in the property panel between the
existing geometry/flow group and the Date Settings group. The field is
deliberately non-scriptable: a scriptable PdmPtrField round-trips in
Python as an opaque "ClassName:0xADDR" string rather than a resolved
object, defeating the point. Two CAF object methods on
RimcPerforationInterval expose the relationship:
- AddFilter (perf.add_filter(filter)) sets the pointer; replaces any
existing filter (single filter per perforation).
- cell_filter (perf.cell_filter()) resolves to the referenced
RimCellFilter object on read, mirroring the valve.template() pattern.
Composition stays in RimCombinedFilter: a script that wants AND/OR
across multiple filters attaches a combined filter as the single
filter on the perforation.
The Python example in PythonExamples builds a case-level AND combined
filter (PERMX 100..20000 AND K=5..10), creates a modeled well path
with two perforation intervals, and shares the same filter between
them. test_perforation_filter.py covers single-set, replace-on-second-
add, and shared-pointer round-trips.
Combined filters used to always show whatever literal string the user
passed in (or the default "Combined Filter"). For programmatic creation
without a name, that string never reflected the children, leaving the
case-level "Data Filters" tree row stale when child bounds changed.
Add an m_autoDeriveName flag (default true) to RimCombinedFilter. When
on, the display name is recomputed on every relevant change — addFilter,
removeFilter, onChildAdded, onChildFilterChanged, and combineMode
toggle — by joining children's name() with " AND " / " OR ". The flag
flips to false the first time the user renames the filter via the UI
property panel or via Python (combined.user_description = "..."),
preserving any explicitly typed name verbatim from then on.
Two child-creation paths needed an extra triggerFilterChanged() so the
combined parent recomputes after the child's final name is set:
addPropertyFilterToCombinedFilter (setDefaults updates name without a
signal) and RimcCombinedFilter_addRangeFilter (Rimc applies user
params after addNewFilter).
The Rimc factory passes setAutoDeriveName(false) only when a name is
supplied. The Python example demonstrates both modes side by side, and
test_combined_filter.py adds three cases: named filter preserves its
name, auto-derived filter tracks bound mutations, and Python rename
locks the auto-derived filter.
RimDataFilterInView previously copied the source filter's name once, in
setSourceFilter() and initAfterRead() via syncNameFromSource(). That
left the wrapper showing the default "Combined Filter" when the source
was renamed after the wrapper was created — typical for the Python
add_combined_filter(name=...) flow, where the wrapper is constructed
during addFilter() before the user-supplied name reaches the source.
Replace the wrapper's reliance on its inherited m_name field for
display with a caf::PdmProxyValueField<QString> that delegates the
read to m_sourceFilter()->name() and the write to setName(). Override
userDescriptionField() so the PDM tree label and property panel route
through the proxy — both directions of rename now propagate.
Matches the m_nameProxy pattern already established in RimCellFilter.
The inherited m_name is no longer the display source of truth but is
left in place so RimNamedObject's serialization stays unchanged.
Expose the case-level RimDataFilterCollection (introduced for
case-shared cell filters) to Python so scripts can build combined
filters with property-filter and IJK-range children without going
through a view.
Scriptable surface:
- RimEclipseCase.m_dataFilterCollection -> case.data_filter_collection()
- RimDataFilterCollection (object + Filters child array) -> filters()
- RimEclipsePropertyFilter (object + m_lowerBound + m_upperBound) ->
pf.lower_bound, pf.upper_bound
- RimCellRangeFilter (object + start_index_{i,j,k}, cell_count_{i,j,k})
New CAF object methods under ProjectDataModelCommands/:
- RimcDataFilterCollection: add_combined_filter(name, combine_mode)
- RimcCombinedFilter: add_property_filter(result_variable, result_type)
and add_range_filter(name, start_{i,j,k}, cell_count_{i,j,k}).
Bounds are not part of add_property_filter because the Python class
generator emits a bare nan literal as the default for double fields
initialised to quiet_NaN, which fails to import. Bounds are set via
the now-scriptable lower_bound/upper_bound fields after creation,
matching the update() pattern used elsewhere in the API.
RicEclipsePropertyFilterFeatureImpl::addPropertyFilterToCombinedFilter
now returns the new RimEclipsePropertyFilter* so the CAF method can
hand it back to Python. setDefaults is relaxed to be view-tolerant:
case-level filters have no Rim3dView ancestor and the eclipse case is
already wired by RimEclipsePropertyFilter::setCase via the owning
collection's onItemsChanged hook.
End-to-end coverage in tests/test_combined_filter.py: round-trip
create, mode toggle, property-filter add/mutate, range-filter
add/mutate. Example script in PythonExamples builds an AND combined
filter that gates PERMX between 100 and 20000 against a K=5..10
slice.
Introduce RimDataFilterCollection on RimEclipseCase to hold filter
configuration that is shared across all views of the case. Each
RimEclipseView gets a RimDataFilterInViewCollection of
RimDataFilterInView wrappers that auto-sync with the case-level list
and expose an independent on/off toggle per view. The wrapper's
applyToCellVisibility delegates to the wrapped RimCellFilter via the
existing unified virtual, so property, range, and combined filters all
work uniformly.
RimDataFilterCollection is built on caf::PdmObjectCollection<RimCellFilter>
and emits a filtersChanged signal on add/remove/reorder/child changes.
The wrapper collection mirrors RimSurfaceInViewCollection's sync pattern:
new wrappers default to checked, removed sources drop their wrappers,
and the wrapper list follows the case order. Per-view wrappers feed into
RivReservoirViewPartMgr::computePropertyVisibility after the existing
view-level property-filter loop.
The RicNew*Feature commands now route into a selected RimDataFilterCollection,
and the existing combined-filter path is generalized to allow combined
filters living under either a per-view RimEclipsePropertyFilterCollection
or the new case-level collection. Several Asserted ancestor lookups in
RimEclipsePropertyFilter, RimCellFilter and RimEclipseResultDefinition
that assumed a Rim3dView ancestor were relaxed to the non-asserting
variant with graceful fallbacks, so case-level filters can be created
and edited without a host view.
Backwards compatible: empty Data Filter collection is a no-op,
existing per-view RimEclipsePropertyFilterCollection and
RimCellFilterCollection paths are unchanged. GeoMech parity deferred.
Replace the dynamic_cast dispatch in RimPolygonCollection::loadData() with a
virtual loadData() on RimPolygonContainer. The base recurses through
subCollections; RimPolygonFile overrides to load file data. The folder
RimPolygonCollection no longer needs its own loadData().
Move RimPolygonFile::canAddSubCollection() and RimPolygonInView::sourceItem()
implementations from headers to .cpp.
Simplify std::unexpected calls in RimcNestedCollectionBase by removing
redundant explicit QString construction.
Port RimSurfaceCollection, RimPolygonCollection and the polygon-in-view
side onto the new caf::PdmNestedCollection framework, replacing the
per-class folder plumbing (m_subCollections field, topmost-folder flag,
add/remove/find sub-collection, factory, "new folder" command) with the
shared base.
- RimSurfaceCollection and RimPolygonCollection now derive from
caf::PdmNestedCollection. Each constructor uses the generic folder
icon and exposes a static createTopmost() that marks the top instance
and applies the branded domain icon; RimOilField calls the factory.
- RimPolygonInViewCollection uses the new RimNestedMirrorCollectionInView
helper, and a new RimPolygonContainer carries the source-side polygon
link.
- RicNewSurfaceCollectionFeature is removed; the generic
RicNewNestedCollectionFeature drives the "Add Folder" action against
both collections.
- Update callers (RimAnnotationCollection, RimEnsembleSurface, polygon
command features) and the scripting wrapper (RimcSurfaceCollection),
plus the Python add-folder binding and polygon test.
Generalise the per-domain "Add Folder" command and the scripting-side
add-folder method so they work against any container that exposes
caf::PdmNestedCollectionBase.
- RicNewNestedCollectionFeature: command feature that walks the
selected PdmObject up to the nearest PdmNestedCollectionBase ancestor
and calls addNewSubCollection(). Hidden via canAddSubCollection() for
leaf containers.
- RimcNestedCollectionBase: scripting wrapper exposing the same
add-sub-collection action through the Python interface, so generated
Python bindings get the operation for free on every derived class.
Neither is wired to a concrete collection yet; the per-domain features
remain in place until the migration commit replaces them.
IconProvider::operator= copied a new resource-string but never cleared
the existing m_pixmap. icon() prefers the pixmap when present, so a
pixmap-based default stamped earlier (e.g. by a base class) would
silently shadow a later resource-string-only assignment in a derived
class.
Reset m_pixmap in operator= so resource-string-only assignments fully
replace pixmap-based ones, and cover the replacement semantics with a
caf::IconProvider regression test: default-overwriting-pixmap,
resource-string-overwriting-pixmap and pixmap-overwriting-pixmap.
Needed for the upcoming nested-collection migration where a base class
stamps a folder pixmap and the topmost instance applies a branded
resource-string icon on top.
Introduce a shared base for tree-shaped folder containers that hold a
list of leaf items together with a list of subcontainers of the same
kind. Replaces the per-class plumbing (subcollections field, top-folder
flag, add/remove/find sub-collection, factory) that is duplicated across
RimSurfaceCollection, RimPolygonCollection and similar candidates.
- caf::PdmNestedCollectionBase: non-templated PdmObject base exposing
collection-name, add-new-subcollection, leaf-check and topmost-folder
marking for generic command features and script methods.
- caf::PdmNestedCollection<SelfT, ItemT>: CRTP template that owns the
typed m_subCollections field and inherits the m_items array from
PdmObjectCollection<ItemT>. Provides allItems(), find-sub-collection,
and a typed factory hook.
- RimNestedMirrorCollectionInView: view-side analogue for in-view
collections that mirror a nested source collection, with orphan-sweep
+ find-or-create + reorder sync.
Design rationale (templated vs non-templated, break-even on derived
class count) lives in docs/pdm-nested-collection-design.md.
No callers migrated in this commit.
The StrEnum-based API (rips/enums.py, generated AppEnum classes) requires
Python 3.11+. Align the docs and copilot-instructions with setup.py.cmake,
which already declares >=3.11.
The property/porosity/data-type arguments accept either a typed StrEnum
(rips.PropertyType.DYNAMIC_NATIVE) or a plain string ("DYNAMIC_NATIVE").
Keep both forms visible to users:
- soil_average_sync.py and input_prop_test_sync.py now use string
arguments and point at their _async counterparts for the typed-enum
form. The sync/async examples otherwise mirror each other, so each
pair documents one form.
- Add test_10k_property_strings to exercise available_properties,
active_cell_property, grid_property, and set_active_cell_property
(including data_type="INTEGER") with strings only, so a future
refactor that narrows the accepted types would fail the suite.
Instance.launch and Instance.find now always either return a connected
Instance or raise RipsError, so their return type narrows from
Optional[Instance] to Instance. In launch(), the unreachable `if pid:`
guard around the Popen result and the trailing `return None` are
removed; the only failure paths raise RipsError.
Update the PythonExamples that previously branched on a None return:
- Drop `if resinsight is None: exit(1)` style guards and `if resinsight
is not None:` wrappers; dedent the body where applicable.
- export_corner_point_grid.py used the None check as a fallback to
launch ResInsight; rewrite as try/except rips.RipsError so the
fallback still works.
- instance_example.py is repurposed as a try/except example for
reporting connection failure.
Extend type annotations across the rest of rips.case beyond the property
and enum methods typed in the previous commit. Adds parameter and return
types to the command wrappers (replace, export_*, create_*, import_*),
the simple getters (time_steps, reservoir_boundingbox,
distance_to_closest_fault, days_since_start, view, views, create_view,
simulation_wells, coarsening_info), the selected-cells and NNC accessors
(selected_cells*, nnc_connections*, nnc_connections_{static,dynamic,
generated}_values*), and the private grpc helpers
(__custom_init__, __request, __generate_property_input_iterator,
__generate_property_input_chunks, __nnc_connections_values_async,
__nnc_values_generator_to_list, __generate_nnc_property_input_chunks).
Uses Iterator[X] for grpc streaming responses, List[X] for collected
sequences, the proto message types from Case_pb2 / Properties_pb2 /
NNCProperties_pb2 / Definitions_pb2 for typed payloads, Optional and
Union[str, List[str]] for nullable or coerced-list parameters, and Any
for the heterogeneous _execute_command return values. No runtime
behavior changes; mypy --strict is clean on case.py.
The previous commit auto-generated StrEnum classes for caf::AppEnum scriptable
fields, but the hand-written gRPC API in rips.case still accepted bare strings
for proto-backed enums (PropertyType, PropertyDataType, NNCPropertyType) and
the AppEnum-backed PorosityModelType.
Add a new rips.enums module with hand-written StrEnum classes for the three
proto-only enums and reuse the auto-generated PorosityModelType from
resinsight_classes. Re-export the new enums from the rips package and narrow
the parameter type hints across cell_count, cell_info_for_active_cells,
available_properties, active_cell_property, selected_cell_property,
grid_property, set_active_cell_property, set_grid_property,
active_cell_centers, active_cell_corners, set_nnc_connections_values, and
grid_property_for_positions (sync and async variants).
Existing string arguments keep working because StrEnum members are strings
and protobuf's EnumTypeWrapper.Value accepts both. A new
test_10k_grid_property_strenum asserts string and enum forms return the same
data.
Auto-generate enum.StrEnum classes from cafPdmPythonGenerator for every
caf::AppEnum-backed scriptable field, deduplicated by C++ AppEnum type.
Generated PdmObject attributes and method arguments now reference the
typed enum (e.g. self.depth_type: DepthType = DepthType.MEASURED_DEPTH)
instead of bare str. Member names are sanitized against Python keywords
and identifier rules (None -> None_, R-G -> R_G).
Existing scripts that pass plain strings keep working because StrEnum
members are strings; the wire format is unchanged.
Bumps python_requires to >=3.11 (StrEnum). Adds an Enum branch in
PdmObjectBase.__convert_to_grpc_value so future IntEnum use is also
handled.
Move the element-type include from each migrated collection header into
its .cpp, leaving only a forward declaration in the header. Follows the
pattern from 9e4d79159 for RimJobCollection, made possible by the
constructor-time PdmObject base check.
Replace the class-level static_assert with a DerivedFromPdmObject concept
checked inside the constructor body. The constructor is lazily instantiated,
so T can be forward-declared at the point PdmObjectCollection<T> is used
as a base class.
Migrated to the templated base, preserving each collection's existing XML
field keyword for project-file backward compatibility:
- RimEnsembleFractureStatisticsCollection
- RimEclipseInputPropertyCollection
- RimPlotDataFilterCollection
- RimGeoMechContourMapViewCollection
- RimIntersectionResultsDefinitionCollection
- RimCustomObjectiveFunctionCollection
- RimStimPlanModelTemplateCollection
- RimEnsembleWellLogsCollection
- RimEclipseContourMapViewCollection
- RimSeismicViewCollection
- RimGridInfoCollection
- RimEclipseStatisticsCaseCollection
- RimWellIASettingsCollection
- RimVfpDataCollection
External callers that accessed the old public/named array fields directly
(RimEclipseInputPropertyCollection::inputProperties,
RimEclipseStatisticsCaseCollection::cases) now use addItem/items() from
the base. RimCustomObjectiveFunction.h replaced its include of the
collection header with a forward declaration to break the new include
cycle introduced by the migration.
RimEnsembleFileSetCollection, RimJobCollection, RimQuickAccessCollection,
RimRefinementRegionCollection and RimFractureTemplateCollection now inherit
from the templated caf::PdmObjectCollection<T> base, reusing the shared
add/insert/delete/items plumbing introduced for RimValveTemplateCollection.
XML field keywords are preserved on every collection so existing project
files continue to load.
Bind a PdmProxyValueField that inverts the Z sign to the Points group in
RimPolygon's property panel, mirroring the pattern already used by
RimExtrudedCurveIntersection. The raw m_pointsInDomainCoords field keeps
its XYZ semantics, scriptable keyword, and serialization unchanged.
Fixes#12106.
The bare "// Overrides" comment is now redundant since the override keyword on
each method already conveys the same information. Qualified comments like
"// Overrides from BaseClass" carry additional information and are kept.
Replace inline bar-to-Pascal magic constants with RiaEclipseUnitTools::barToPascal
for initial stress and pore pressure parameters passed to the Well Impact Analysis
exporter. Behavior is unchanged; both sites previously multiplied by 100000.0.
Fixes#5840.
isFaceNormalsOutwards lazily computes whether cell face normals point outward
and caches the answer in m_isFaceNormalsOutwardsComputed. setFlipAxis was
mirroring the node coordinates without resetting that cache, so any consumer
that had triggered the lazy computation before the user toggled Flip X/Y kept
the pre-flip answer. The well-path/cell intersection in
RigEclipseWellLogExtractor::calculateIntersection then inverted the
entering/leaving flag the wrong way, causing the proper-pair filter in
populateReturnArrays to discard every intermediate cell along a perforation
interval - only the start and end cells survived through the
well-starts/ends-inside-a-cell fallback.
Reset m_isFaceNormalsOutwardsComputed when the flip state actually changes.
Add unit tests covering all four combinations of Flip X / Flip Y for both the
data-level node mirroring and the well-path/cell intersection (freshly-built
grid and interactive-toggle scenarios).
When a project file persisted FlipXAxis or FlipYAxis as true, the flag was
restored to the PdmField but never propagated to the grid nodes during
RimEclipseResultCase::importGridAndResultMetaData. Only the UI change handler
applied the flip, so on reload the grid stayed un-flipped while the checkbox
appeared active.
Mirror the call already present in RimEclipseInputCase and RimRoffCase by
invoking setFlipAxis before computeCachedData.
Remove the hard floor against m_defaultPerspectiveNearPlaneDistance
(0.05) when computing the perspective near plane from the closest
bounding-box corner. For very small models the closest corner can sit
within that distance, and clamping pushed the near plane past the
model so it was clipped entirely; parallel projection was unaffected
because its near plane scales directly with the bbox.
The zero/negative guard at the end of the block is the actual safety
net, and the maxFarNearRatio constraint still protects depth-buffer
precision when geometry spans both very close and very distant
corners.
Expose XYZ extent and XY UTM offset as user-editable fields in the
Customize Mock Model dialog instead of hard-coding them in
RimEclipseResultCase. Defaults match the previous hard-coded values
(6000/12000/500 extent, 400000/6000000 offset).
Switch the message parameter from const QString& to std::string_view so
callers with std::string in hand no longer need a QString::fromStdString
wrapper. The internal duration formatting still uses QString::arg, fed
from a single QString::fromUtf8 conversion.
Callers passing QString locals are updated with .toStdString() at the
call site.
Move the QMessageBox-displaying helper out of the logging utility into a
dedicated UI class. RiaLogging.h no longer pulls QMessageBox/QWidget into
the ~270 files that include it. Behavior is preserved: showError displays
the dialog when running interactively (and not under regression tests),
then forwards the text to RiaLogging::error.
Restore m_filePathInCache_OBSOLETE as a read-only field (write disabled)
so existing projects that stored the well path in WellPathFilePathInCache
can still be loaded. In initAfterRead, look up the cache file in the
current project's cache directory and copy the path into m_filePath.
Wrap the discarded calls to the [[nodiscard]] std::expected-returning
transformKeywordsInDeckFile in EXPECT_TRUE(...has_value()) to silence
C4834 and assert that the transformation succeeded.
Robustness fixes in RivPipeGeometryGenerator:
- updateFilteredPipeCenterCoords: drop NaN/Inf input coords. The main loop
now skips iterations where any of the three sampled coords is undefined,
and the trailing endpoint walks back to the last finite coord so the
segment-to-result bookkeeping invariant holds.
- computeExtrudedCoordsAndNormals: stop asserting on parallel plane/ray
intersections (these can occur at very sharp bends due to numerical
issues) and return early when the extruded node count does not match the
expected cross-section node count.
Magne Sjaastad
Kristian Bendiksen
copilot-swe-agent[bot]andmagnesj