Stop format_name_template from mutating the caller substitutions dict

[chuenchen309]

Describe changes

format_name_template mutates the caller's substitutions dict.

substitutions or {} only builds a new dict when the argument is falsy, so a non-empty dict is aliased and then written into by the .update() / .setdefault() calls below:

subs = {"custom": "abc"}
format_name_template("run-{custom}-{date}-{time}", substitutions=subs)
subs
# {'custom': 'abc', 'date': '2026_07_16', 'time': '16_50_23_522789'}

An empty dict masks the bug ({} or {} yields a fresh dict), which is likely why it went unnoticed.

Why it matters

_create_snapshot passes pipeline_configuration.substitutions straight in when deriving a schedule name (pipeline_definition.py:995-998), and that happens before the snapshot is serialized into the PipelineSnapshotRequest (:1085) — so the injected date/time get persisted into the snapshot.

finalize_substitutions then uses setdefault, so the baked-in values win over the real start time from that point on:

cfg = PipelineConfiguration(name="p", substitutions={"env": "prod"})
format_name_template("p-{env}-{date}", substitutions=cfg.substitutions)

cfg.finalize_substitutions(start_time=datetime(2027, 1, 1, 9, 0, 0), inplace=False)["date"]
# '2026_07_16'   <- the future start time is ignored

Every run off that schedule then resolves the same frozen {date}/{time}, i.e. the same run name — the collision the server warns about in sql_zen_store.py. Triggering it needs nothing unusual: a schedule without an explicit name, a non-empty @pipeline(substitutions={...}) (public, documented), and a run_name_template containing {date}/{time} — which is the default (run_utils.py:73).

Changes

One line: copy the dict. Explicit date/time substitutions still take precedence, and the returned name is unchanged.

Two tests added — the mutation guard fails on develop and passes here; the second pins that caller-supplied date/time are still honored. format_name_template had no test coverage, which is why this survived.

Pre-requisites

• I have read the CONTRIBUTING.md document.
• I have added tests to cover my changes.
• I have based my new branch on develop and the open PR is targeting develop.
• IMPORTANT: I made sure that my changes are reflected properly in the following resources:
  • ZenML Docs — internal helper, no doc change needed
  • Dashboard: not affected
  • Templates: not affected
  • Projects: not affected

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Other (add details above)

Verification

• pytest tests/unit/utils/test_string_utils.py — 6 passed.
• ruff check (with --extend-ignore D for tests, per scripts/lint.sh) and ruff format --check clean.
• mypy src/zenml/utils/string_utils.py reports one pre-existing unused-ignore at line 252, unrelated to this change — it appears identically on a clean develop (verified via git stash).

xref: #5080 also touches string_utils.py, but only adds a new append_random_suffix helper near the top of the file — no overlap with format_name_template.

I could not add a release-notes label myself (external contributors lack label permissions); this looks like no-release-notes to me, but happy to defer.

---

AI disclosure: drafted with Claude Code (Opus 4.8), including the root-cause trace and the tests. I ran the repro and the tests locally and reviewed the diff before opening.

[chuenchen309]

Closing as a duplicate — this is the same fix as #5088, which I opened earlier today and had lost track of across parallel work. Apologies for the noise; #5088 stands.

Sorry for the extra notification.