⬆️ Bump anaconda from 2026.06 to 2026.07

[mmcky]

Rolling monthly bump of the conda env pin anaconda=2026.06 → 2026.07.

Why this is hand-authored rather than a Dependabot PR

Dependabot assigns each repo a fixed weekly slot, and this repo's slot had not yet fired for the 2026.07 release. The rest of the Python lecture family has already been bumped, validated and merged, so this opens the PR by hand to finish the sweep rather than waiting several more days.

If Dependabot later opens its own PR for the same bump it will not recognise this one, so that duplicate should be closed (as happened with lecture-python-advanced.myst#349 during the 2026.06 round).

Validation

The pull_request CI check is not a meaningful test of an env-only bump — it downloads the main build-cache artifact and reuses the cached notebook outputs, so it skips re-executing the notebooks under the new env entirely. A green check there proves nothing about the upgrade.

The canonical signal is a cache.yml dispatch on this branch, which does a full fresh jb build lectures -W --keep-going (warnings→errors) executing every notebook under 2026.07. I have dispatched that run and will report the result here.

Known 2026.07 breakage — pre-checked

The one regression found so far in this sweep is that anaconda 2026.07's matplotlib removes the long-deprecated matplotlib.cm.get_cmap, which raises ImportError at import time and kills the whole notebook (hit lecture-python-intro laffer_adaptive.md, fixed in QuantEcon/lecture-python-intro#798). I grepped this repo's lectures/ for that pattern before opening — no hits.

Cross-repo status for this sweep is tracked in QuantEcon/workspace-lectures#21.

[Copilot]

Pull request overview

This PR performs the monthly roll-forward of the conda environment pin for the lecture build environment by updating the anaconda metapackage version in environment.yml, keeping the repo aligned with the rest of the lecture-family environment sweep.

Changes:

• Bump anaconda from 2026.06 to 2026.07 in the conda environment specification.

[github-actions]

🚀 Deployed on https://6a5d84b3723978ed9de8d08d--epic-agnesi-957267.netlify.app

[mmcky]

✅ Validated — safe to merge.

Validated by dispatching cache.yml on this branch, which does a full fresh jb build lectures -W --keep-going (warnings→errors), executing every notebook under anaconda 2026.07 rather than reusing the main build cache.

Result: run 29708297855 completed success, build job 6m37s. That is in line with this repo's normal cache-build time (the last four scheduled runs on main took 7m46s, 8m07s, 7m27s and 7m17s; the equivalent 2026.06 validation took 7m36s), so it is a full execution pass rather than a short-circuit.

I also grepped lectures/ beforehand for the one regression this sweep has turned up — anaconda 2026.07's matplotlib removes the long-deprecated matplotlib.cm.get_cmap, which raises ImportError at import time (it broke lecture-python-intro laffer_adaptive.md, fixed in QuantEcon/lecture-python-intro#798). No hits in this repo.

Note on the delay: this run sat queued for ~95 minutes before starting, and its total wall-clock of 164m reflects that rather than build time. Both RunsOn jobs on this branch (cache.yml and ci.yml) were stuck while the GitHub-hosted job on the same commit ran normally — GitHub opened an Actions incident at 23:34 UTC, six minutes before these were dispatched, and the workflow_job.queued events that RunsOn relies on to provision the g4dn.2xlarge were evidently delayed. RunsOn picked the job up by itself once the backlog drained, with no re-dispatch and no configuration change. Nothing here indicates an AWS-capacity or spot=false problem.

Note: the pull_request CI check isn't a meaningful test of an env-only bump — it downloads the main build-cache artifact and skips re-executing unchanged notebooks. The cache-workflow run above is the canonical signal.

This completes the 2026.07 sweep across the Python lecture family — cross-repo status is tracked in QuantEcon/workspace-lectures#21.