Discrete variables support in GEMSEO - #92
Open
FrancoisGallard wants to merge 21 commits into
Open
FrancoisGallard wants to merge 21 commits into
FrancoisGallard wants to merge 21 commits into
Conversation
Add philote_mdo.gemseo, providing PhiloteDiscipline (a GEMSEO Discipline that calls a remote Philote-MDO server) and GEMSEOtoPhiloteDiscipline (the opposite direction, serving a GEMSEO discipline over gRPC). Add matching examples (Paraboloid, OpenAeroStruct, and the Sellar problem driven from OpenMDAO) and integration tests under tests/, adapted to the current OpenMdaoSubProblem() + add_group() construction pattern. Also add the gemseo extra to pyproject.toml, add gemseo to the CI test dependencies, and fix proto compilation on Windows.
Add a "Working with GEMSEO" section with two tutorials, ported from the retired Jupyter Book docs: using PhiloteDiscipline to optimize a remote Paraboloid discipline, and coupling OpenAeroStruct with GEMSEO through an OpenMDAO sub-problem. Enable @docusaurus/theme-mermaid so the flowchart diagrams on these pages render. Note: docs/package-lock.json was not regenerated (no Node/npm available in this environment) -- run `npm install` in docs/ before the next `npm ci` build.
version in pyproject.toml
Covers the ValueError on a missing channel, sending discipline options to the server, the ndarray-sized default_output_data branch in GEMSEOtoPhiloteDiscipline.setup(), and the defensive skip of Jacobian entries for outputs outside a wrapped discipline's own output grammar. Also marks the TYPE_CHECKING-only imports as pragma: no cover, since they can never execute at runtime. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
…nder The GEMSEO tutorials use ```mermaid fences, but the theme was never actually wired in: package.json didn't list the dependency and docusaurus.config.ts had no markdown.mermaid / themes entry, so the diagrams were not compiled and rendered as plain code blocks instead. Verified by building the site and loading both tutorial pages in a browser: the flowcharts now render as SVG diagrams. Note: docs/package-lock.json still needs to be regenerated with real npm (this environment only has bun) before `npm ci` in CI will pick up the new dependency. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The 0.8.0 versioned docs describe the released package, which has no philote_mdo.gemseo module. The GEMSEO pages stay under docs/docs/ (the Next version) and will be snapshotted by the release workflow with the first release that ships them. Signed-off-by: Christopher Lupp <christopherlupp@gmail.com>
package.json gained @docusaurus/theme-mermaid without a matching lock update, so the documentation workflow's npm ci refused to install. Regenerated with npm 10 (the version CI's Node 20 ships) in lockfile-only mode: this adds mermaid and its dependencies, and bumps katex from 0.16.45 to 0.16.47 because mermaid requires ^0.16.47. Signed-off-by: Christopher Lupp <christopherlupp@gmail.com>
philote-examples 0.5.1 is the first release whose OasAerostructDiscipline declares the CD and CL partials with respect to alpha. With earlier versions the SLSQP scenario fails on its first gradient request. Signed-off-by: Christopher Lupp <christopherlupp@gmail.com>
PhiloteDiscipline is built on ExplicitClient, so it connects to explicit Philote-MDO servers only; against an implicit server it fails at execute with "Method not found". The intro also named RemoteExplicitComponent, the OpenMDAO client, as the way to serve an OpenMDAO model; serving is done with OpenMdaoSubProblem. Signed-off-by: Christopher Lupp <christopherlupp@gmail.com>
The GEMSEO tests import gemseo at module level, so without it installed they fail to import. List it next to OpenMDAO in the README and the installation page. Signed-off-by: Christopher Lupp <christopherlupp@gmail.com>
Fix the misspelled GEMSEOtoPhiloteDiscipline attribute and constructor argument before they are released as public API. Signed-off-by: Christopher Lupp <christopherlupp@gmail.com>
The constructor documented that an empty name would use the remote discipline's name, but passed it straight to GEMSEO, so every remote discipline was named PhiloteDiscipline. Query GetInfo and use the server-reported name when none is given; GEMSEO still falls back to the class name when the server reports none (the default). Dynamic-shape and discrete server variables were not handled. Dynamic shapes were never sent, so the first execute failed with an error asking for SetVariableShapes, which a GEMSEO user cannot call. Discrete variables were left out of the grammars and silently kept their server-side defaults. Reject both at construction with a NotImplementedError naming the variables, until they are supported. Signed-off-by: Christopher Lupp <christopherlupp@gmail.com>
test_sellar_mda_compute only asserted that obj and con1 were present, so any values passed. Evaluate at the canonical Sellar starting point (x = 1, z = [5, 2]) and compare against the canonical result there. Signed-off-by: Christopher Lupp <christopherlupp@gmail.com>
Signed-off-by: Christopher Lupp <christopherlupp@gmail.com>
Python 3.9 reached end of life in October 2025, and GEMSEO 6.3 requires 3.10, so the 3.9 CI job resolved GEMSEO 6.2 while every other job tested 6.3. Require Python >=3.10, remove 3.9 from the CI matrix and the classifiers, and update the agent guidelines and the landing page. Signed-off-by: Christopher Lupp <christopherlupp@gmail.com>
PhiloteDiscipline rejected any server that declared discrete variables. They are now part of the GEMSEO grammars next to the continuous ones, bound to no type, since a Philote discrete variable carries a google.protobuf.Value and may hold any JSON-compatible value. Discrete inputs are split off the GEMSEO input data and sent on their own side of the wire protocol, for both ComputeFunction and ComputeGradient, and the discrete outputs returned by the server are merged back into the output data. Discrete variables are excluded from the Jacobian. GEMSEO already filters non-continuous variables out when the differentiated variables are chosen with add_differentiated_inputs/outputs, but not when the whole Jacobian is requested, so _get_differentiated_io is overridden to do it there too. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Every variable of the wrapped GEMSEO discipline was served as a flat continuous Philote variable. The names of the input and output grammars are now classified with their data converter's is_numeric(): a numeric name is declared as a continuous Philote variable, any other name as a discrete one, which carries any JSON-compatible value. The discrete inputs sent by the client are merged back into the GEMSEO input data before execute() and linearize(), and the discrete outputs of the wrapped discipline are streamed back to the client. Only the variables holding continuous data take part in the Jacobian, so that no partial derivative is declared that compute_partials() would then leave at zero. This also drops the integer-valued variables, which are numeric but not differentiable. compute_partials() therefore selects the differentiated variables with add_differentiated_inputs/outputs, which apply that same filter, rather than asking for the whole Jacobian. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Add a "Discrete variables" section to the GEMSEO page covering both directions: how GEMSEOtoPhiloteDiscipline classifies a grammar name with its data converter's is_numeric(), and how PhiloteDiscipline puts the server's discrete variables in the GEMSEO grammars. Note that they are never differentiated, and that Philote-MDO does not transfer their default values. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Two branches of GEMSEOtoPhiloteDiscipline were unreached: skipping an integer-valued output in setup_partials(), and the early return of compute_partials() when nothing was declared to differentiate. Add a discipline with integer variables and one whose variables are all non-numeric, which brings philote_mdo.gemseo to 100% statement and branch coverage. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…umeric A continuous Philote variable travels as an array of doubles, and the DiscreteVariable message carries a google.protobuf.Value whose documented contents include integers. An integer-valued variable therefore belongs to the discrete side of the protocol, as it does in OpenMDAO, whose bindings map Philote discrete variables onto add_discrete_input/output. is_numeric differs from is_continuous by exactly the int type, so it sent integers down the double path, where they were silently converted to floats and came back as floats. This also removes the need for the second predicate: setup_partials() had to re-filter its declarations with is_continuous so as not to declare a partial derivative that compute_partials() would leave at zero. The classification and the Jacobian now agree by construction. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add discrete variables support in PhiloteDiscipline and GEMSEOtoPhiloteDiscipline