Limit search space to relevant technologies

[tsmbland]

Description

Adding some code to initialize_from_technologies, which initializes the search space, to only consider technologies that produce at least one demanded commodity. Ultimately, if there are any technologies in the sector that do not produce any demanded commodities, removing them from the search space will shrink the size of the linear problem at no cost.

This will be particularly important once I've done #684 and users can split commodity demands across multiple subsectors. This should then be a viable solution for fixing the memory problems in #389 (watch this space...).

I considered making this a separate filter, but I want it to be included in all models, for all agents, regardless of what filters the user has put in their agents file, so this seemed like the most appropriate place to put it.

Added a test to show that it works.

There was one stubbornly failing test (test_run_retro_agent), which I've managed to keep happy in the end. Not too fussed about it though as it's to do with retrofit agents which are on their way out anyway

[Copilot]

Pull Request Overview

This PR limits the search space in the model to only include technologies that produce at least one demanded commodity, thereby reducing the size of the underlying mathematical problem.

• Added filtering logic in initialize_from_technologies in src/muse/filters.py.
• Updated tests in tests/test_filters.py to verify correct behavior when no demanded commodities are produced.
• Modified tests in tests/test_agents.py to account for region-specific selection when invoking retro_agent.next.

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
tests/test_filters.py	Added tests to ensure search_space correctly reflects commodity demand.
tests/test_agents.py	Updated retro_agent.next call to perform selection by region.
src/muse/filters.py	Introduced filtering logic to combine search_space with demanded commodity mask.

Comments suppressed due to low confidence (2)

src/muse/filters.py:456

• [nitpick] Ensure that the boolean mask 'produces_demanded_commodity' correctly broadcasts to all dimensions of 'search_space'. Including an inline comment or an assert to document the expected dimensional consistency could help prevent subtle filtering issues.

return search_space & produces_demanded_commodity

tests/test_agents.py:147

• [nitpick] Consider adding a comment to clarify the rationale behind hardcoding the region selection to index 0 or, if applicable, extend the test data to cover scenarios with multiple regions.

retro_agent.next(technologies.sel(year=investment_year).isel(region=0), agent_market.isel(region=0), demand_share,)

[dalonsoa]

I think it makes perfect sense the way you implemented it. It is surprising this was not already in place - why investing in a technology that does not produce any demanded commodity?