[perf-scan] 2026-08-15 -- performance opportunities found

[github-actions]

Performance Scan - 2026-08-15

Automated scan of src/apm_cli/ for algorithmic performance anti-patterns.
2 finding(s) identified.

Findings

[A] Double-pass over refs in sort_remote_refs -- deps/git_remote_ops.py:105-111

• Current: O(2n) -- refs list iterated twice via separate list comprehensions to partition tags and branches; called once per dependency resolution (git-hosted deps)
• Proposed: O(n) -- single-pass loop partitions refs into tags and branches simultaneously
• Fix: Replace the two comprehensions at lines 107-108 with a single for r in refs loop that appends to tags or branches based on r.ref_type, eliminating the second full traversal.

[E] Full install stack eagerly imported on every CLI invocation -- cli/cli.py:27-56

• Current: cli.py imports all 20+ command modules (including commands/install.py) unconditionally at startup; commands/install.py transitively pulls in the entire install subtree (~18 000 lines across 30+ modules); lightweight subcommands such as apm config, apm doctor, and apm list bear the full import cost every run
• Proposed: Lazy command-module loading so each commands/*.py file is imported only when its subcommand is actually invoked
• Fix: Replace the flat module-level imports in cli.py with a Click lazy-loading group (e.g. click.Group subclass that defers __import__ per command) so that apm config get key does not load the install, marketplace, or compile stacks.

Scan coverage

• src/apm_cli/ (430 files scanned)
• Patterns checked: A (quadratic loops), B (linear scan in loop),
  C (unconditional expensive ops), D (redundant config parsing),
  E (heavy top-level imports), F (sequential independent I/O)

Generated by Daily Performance Scanner · 121 AIC · ⌖ 6.22 AIC · ⊞ 7.3K · ◷

• expires on Aug 22, 2026, 1:43 AM UTC