Print friendly plugin names, and print both plugins' names for conflicting plugins

[agriyakhetarpal]

Description

In the CI for a project at wntrblm/nox#1095, I recently ran into the following error:

PluginError: Conflicting plugins found for `post_solves`:

  - CondaPostSolve(name='signature-verification', action=<conda.trust.signature_verification._SignatureVerification object at 0x10774b770>)

Multiple conda plugins are registered via the `conda_post_solves` hook. Please make sure that you don't have any incompatible plugins installed.

The reason for this is that the CI setup uses conda 25.11.1 with setup-miniconda@v3. That means there is an older version of conda that still implements the signature-verification post_solves hook, and a newer conda-content-trust that does the same. In this scenario, run_constrained: conda-content-trust >=0.3.0 was added only with conda 26.3.

After talking to @kenodegard and @jaimergp, we can do two things, which this PR does:

1. Use importlib.metadata.packages_distributions to output friendly plugin names. We fall back to the raw pluggy plugin name in case this isn't possible, which may happen when plugin developers perform editable installations, and they should understand a module path just fine. However, this won't be the case for a majority of plugin users, of course.
2. Restructure the conflict detection for plugins from a flat list to one where we print out plugins grouped by name, so both sides of each conflict appear on the same line.

Therefore, the old output was - CondaPostSolve(name='signature-verification', action=<...object...>), and the new output should be something along the lines of - 'signature-verification' provided by: split on one line per provider, like this (for example):

Conflicting plugins found for `<hook>`:
  - 'archspec' provided by:
    - conda 26.1.0
    - conda-content-trust 0.2.0

Checklist - did you ...

• Add a file to the news directory (using the template) for the next release's release notes?
• Add / update necessary tests?
• Add / update outdated documentation?

[codspeed-hq]

Merging this PR will not alter performance

✅ 35 untouched benchmarks
⏩ 25 skipped benchmarks¹

---

_{Comparing agriyakhetarpal:conda-plugins-conflict (06bf417) with main (4572fe6)}

Footnotes

1. 25 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[agriyakhetarpal]

The test failures are unrelated; it looks like the tests are getting cancelled for some reason. Thanks!

[kenodegard]

nice followup to the impl.plugin_name work from #15840

there are probably other places adding this distname+version would also be useful (e.g., conda info)

[agriyakhetarpal]

Done in cb8cf6a and 76473fc. There are some other uses of plugin names, but these two are the only ones that are applicable in this context.

[kenodegard]

Let's also sort the list of conflicts:

Suggested change

	[_plugin_source(p.impl.plugin_name) for p in conflicting],
	[_plugin_source(p.impl.plugin_name) for p in sorted(conflicting)],

[agriyakhetarpal]

Thanks! Noting here that CondaPlugin dataclasses don't have an lt dunder, so I had to do sorted(conflicting, key=lambda p: p.impl.plugin_name) for this, to sort by the string plugin name.

[jezdez]

I opened agriyakhetarpal#5 as a concrete sketch for the plugin-manager-method version of this change.

[agriyakhetarpal]

I think we might need to harden get_canonical_name a bit so that we can be more helpful:

from conda.base.context import context
from conda.plugins.hookspec import hookimpl
from conda.plugins.types import CondaSolver
from conda.exceptions import PluginError

pm = context.plugin_manager

class IntruderSolver:
    @hookimpl
    def conda_solvers(self):
        yield CondaSolver(name='classic', backend=object)

pm.register(IntruderSolver())

try:
    pm.get_hook_results('solvers')
except PluginError as e:
    print(str(e))

which gets us

Conflicting plugins found for `solvers`:

  - 'classic' provided by:
    - <unknown_module>.IntruderSolver[4392487968]
    - conda.plugins.solvers

Multiple conda plugins are registered via the `conda_solvers` hook. Please make sure that you don't have any incompatible plugins installed.

[jezdez]

I opened agriyakhetarpal#6 as a follow-up for the remaining review tweaks: hardening canonical names for registered plugin instances, tightening the conflict-message tests, and clarifying the entry point bookkeeping/news wording.

[github-actions]

Bencher Report

Branch	conda-plugins-conflict
Testbed	ubuntu-22.04

Click to view all benchmark results

Benchmark
tests/cli/test_startup_benchmarks.py::test_context_init
tests/cli/test_startup_benchmarks.py::test_generate_parser
tests/cli/test_startup_benchmarks.py::test_import_cli_main
tests/cli/test_startup_benchmarks.py::test_import_conda_argparse
tests/cli/test_startup_benchmarks.py::test_import_context
tests/cli/test_startup_benchmarks.py::test_version_main
tests/cli/test_subcommands.py::test_env_create[classic]
tests/cli/test_subcommands.py::test_env_create[libmamba]
tests/cli/test_subcommands.py::test_env_list_benchmark[classic]
tests/cli/test_subcommands.py::test_env_list_benchmark[libmamba]
tests/cli/test_subcommands.py::test_env_update[classic]
tests/cli/test_subcommands.py::test_env_update[libmamba]
tests/cli/test_subcommands.py::test_install[classic]
tests/cli/test_subcommands.py::test_install[libmamba]
tests/cli/test_subcommands.py::test_list[classic]
tests/cli/test_subcommands.py::test_list[libmamba]
tests/cli/test_subcommands.py::test_run[classic]
tests/cli/test_subcommands.py::test_run[libmamba]
tests/cli/test_subcommands.py::test_update[classic-update]
tests/cli/test_subcommands.py::test_update[classic-upgrade]
tests/cli/test_subcommands.py::test_update[libmamba-update]
tests/cli/test_subcommands.py::test_update[libmamba-upgrade]
tests/core/test_solve.py::test_solve_1[classic]
tests/core/test_solve.py::test_solve_1[libmamba]
tests/models/test_match_spec.py::test_match_1
tests/shards/test_shards_subset.py::test_traversal_algorithm_benchmarks[devops_automation-bfs-cold]
tests/shards/test_shards_subset.py::test_traversal_algorithm_benchmarks[devops_automation-bfs-warm]
tests/shards/test_shards_subset.py::test_traversal_algorithm_benchmarks[devops_automation-pipelined-cold]
tests/shards/test_shards_subset.py::test_traversal_algorithm_benchmarks[devops_automation-pipelined-warm]
tests/shards/test_shards_subset.py::test_traversal_algorithm_benchmarks[python-bfs-cold]
tests/shards/test_shards_subset.py::test_traversal_algorithm_benchmarks[python-bfs-warm]
tests/shards/test_shards_subset.py::test_traversal_algorithm_benchmarks[python-pipelined-cold]
tests/shards/test_shards_subset.py::test_traversal_algorithm_benchmarks[python-pipelined-warm]

🐰 View full continuous benchmarking report in Bencher

[agriyakhetarpal]

Thanks @kenodegard, makes sense to me!