If some internal attributes are saved in the templates, their description
should be checked just like the input attributes to ensure there are no
conflicts.
For every node, the only information that is kept when saving a
template are the non-default input attributes' values as well as the
node's position.
The unit test checking the templates' nodes is updated to stop taking
outputs into account.
This test checks that there is no compatibility issue with the nodes
saved in the template files. It fails when an upgrade of a templates is
needed. Any template can still be opened even if its nodes are not up-to-
date, as they will be automatically upgraded.
Duplicates used to be stored in a dictionary with an entry being
"parent node": "duplicated node". On occasions where a single
parent node was duplicated more than once, the latest duplicated
node erased the previous one(s), and these older ones were
"lost": after being created, there was no trace left of their
existence in the duplication operation. Undoing that duplication
operation was thus leaving these duplicated nodes out and not
removing them.
Duplicated nodes are now stored as "parent node": [list of
duplicated nodes] to keep track of all the created nodes,
effectively removing them upon an "undo".
- Support loading external pipelines as templates with project files
- Add template files for some standard pipelines
- Remove the hard-coded generation of new pipelines
- Update multiviewPipeline test: the multiviewPipeline test relied on the hard-coded generation of
pipelines.
* add 'matchDescription' method on desc.Attribute.
* desc.ListAttribute/desc.GroupAttribute: 'matchDescription' ensure that value perfectly match element/group description. Enable recursive checking for child attributes.
* nodeFactory: detect DescriptionConflicts in pre-process pass by checking that serialized node's attributes perfectly match their descriptions
* CompatibilityNode.attributeDescFromNameAttribute: consider link expressions as valid values
* test_compatibility: more complete description conflicts unit testing
CompatibilityNodes introduce a new notion of computability per node.
An uncomputed (and therefore uncomputable) CompatibilityNode blocks the computation of all its successor.
* update nodes computability on topology change (in addition to min/max depth)
* evaluate leaves computability to determine if the whole graph can be processed
handle this low-level operation engine side
* fix ListAttribute children links duplication
* handle CompatibilityNode duplication
* move corresponding unit test in test_graph.py
* [ui] add DuplicateNodeCommand
Improve node serialization/deserialization to be able to recreate the exact same node in the graph when loading a meshroom project, even if the corresponding node's description has changed or does not exist anymore. This allows to recover already computed data on disk, without being impacted by changed uids. CompatibilityNode also provides an on-demand upgrade system to turn into a Node that meets the current node description (if possible).
* new abstract class BaseNode, base class for Node and CompatibiliyNode
* Node: serialize everything needed to spawn a CompatibilityNode with precomputed outputs: inputs, uids, parallelization settings, unresolved internal folders and outputs
* node_factory: handles node deserialization and compatibility issues to create either a Node or a CompatibilityNode
* add compatibility unit tests
* add 'sfmPipeline', 'mvsPipeline' and 'photogrammetryPipeline' instantiating nodes in an existing graph
* rename main function (creating a new Graph) 'photogrammetry' + create a Publish node if 'output' is specified
* Add chunk notion for parallelization
* Allows Node desc to implement custom updateInternals
* CameraInit node implement a specific updateInternals to update the
input image list
* FeatureExtraction, FeatureMatching, DepthMap, DepthMapFilter:
implement parallelization
Inputs and outputs Attributes are now declared in two separate lists
* no more isOutput field on Attribute description
* guarantee attributes order (was not the case in Python2 with class attributes)
* Attribute's name is now part of its description
Note: for now, input and output Attributes are still stored in a single model inside a Graph.
* ensure that identical graphs have the same UIDs
* ensure that multiview pipeline graphs with different input images have
different UIDs on all output attributes
- flowEdges: Compute a transitive reduction of the graph to provide
node-level minimal dependencies to create an execution workflow for
renderfarm.
- minMaxDepthPerNode: compute the min and max depth per node
