cellpy 2.1.3.post3__tar.gz → 2.1.4__tar.gz

cellpy-2.1.4/.issueflows/03-solved-issues/cycle_status_2026-09-05.md

@@ -0,0 +1,27 @@
+# Cycle status
+
+- Queue spec: `yolo` (resolved `label:yolo`)
+- Repo: `jepegit/cellpy` (`C:\scripting\cellpy-workspace\cellpy`), default branch `master`
+- Failure policy: `onfail:stop`
+- Started: 2026-09-05T19:25:00+02:00
+- Finished: 2026-09-05T22:20:00+02:00
+- Confirmed: yes (single consolidated confirm, 3 issues)
+
+- [x] Done
+
+## Queue (ordered)
+
+- [x] #990 — cellpy new: honour no_input when the project directory does not exist — merged https://github.com/jepegit/cellpy/pull/994
+- [x] #991 — Add cli_api.list_templates() returning the batch templates as data — merged https://github.com/jepegit/cellpy/pull/995
+- [x] #993 — Docstring cross-references lost their module paths in #968 — merged https://github.com/jepegit/cellpy/pull/996
+
+Blocked: none. Skipped (closed): none. No stop condition tripped.
+
+## Notes
+
+- Leftover `issue985_*` group (`- [ ] Done`) was swept from `01-current-issues/`
+  to `02-partly-solved-issues/` by the first `/iflow-init`.
+- Every PR needed one `gh pr checks --watch` pass before `gh pr merge --squash`
+  succeeded (base branch policy requires the `essential` + `full` checks).
+- Local branches `990-no-input-project-dir`, `991-list-templates-data`,
+  `993-dotted-docstring-refs` are left for `/iflow-cleanup`.

cellpy-2.1.4/.issueflows/03-solved-issues/cycle_status_2026-09-07.md

@@ -0,0 +1,22 @@
+# Cycle status
+
+- queue: yolo
+- resolved: label:yolo
+- repo: jepegit/cellpy
+- onfail: stop
+- started: 2026-09-07T19:33:00Z
+- stopped: 2026-09-07T20:10:00Z
+
+## Queue
+
+- [x] #937 — Notebook tooling (ipykernel, matplotlib) is a hard runtime dependency — ~90 MB in a headless server image — merged https://github.com/jepegit/cellpy/pull/1002
+- [~] #938 — Missing external tools fail silently: mdb-export raises bare FileNotFoundError, pyodbc ImportError hides two loaders — failed: not yolo-small (three independent deliverables; see 02-partly-solved-issues)
+- [ ] #960 — Possible bugs in cellpy setup and configuration — not reached
+- [ ] #982 — Default group name from cellpy_db — not reached
+- [ ] #1000 — prepare for changes in batch journal json file — not reached
+
+## Stop reason
+
+Yolo scope check on #938 aborted: issue body is two features (mdb-export typed error + `list_instruments` availability API) plus an owner comment adding a third (`examplesdir` default). Cycle `onfail:stop`. Branch `938-missing-external-tools` holds the capture.
+
+- [x] Done

cellpy-2.1.4/.issueflows/03-solved-issues/cycle_status_2026-09-08.md

@@ -0,0 +1,28 @@
+# Cycle status
+
+- queue: yolo (resolved `label:yolo`)
+- repo: jepegit/cellpy
+- onfail: stop
+- started: 2026-09-08T06:58:00+02:00
+- stopped: 2026-09-08T08:45:00+02:00
+
+## Queue
+
+- [x] #960 — Possible bugs in cellpy setup and configuration — merged https://github.com/jepegit/cellpy/pull/1004
+- [x] #982 — Default group name from cellpy_db — merged https://github.com/jepegit/cellpy/pull/1005
+- [x] #1000 — prepare for changes in batch journal json file — merged https://github.com/jepegit/cellpy/pull/1006
+
+blocked: none
+skipped: none
+
+## Result
+
+All three queued issues went through the full yolo chain and merged. Cycle never halted.
+
+- #960 → https://github.com/jepegit/cellpy/pull/1004
+- #982 → https://github.com/jepegit/cellpy/pull/1005
+- #1000 → https://github.com/jepegit/cellpy/pull/1006
+
+Local close of #1000 hit dirty `AGENTS.md` (unrelated indent). Discarded, then `git switch master` + `git pull --ff-only` landed squash `3d199790`.
+
+- [x] Done

cellpy-2.1.4/.issueflows/03-solved-issues/issue1000_original.md

@@ -0,0 +1,9 @@
+# Issue #1000: prepare for changes in batch journal json file
+
+Source: https://github.com/jepegit/cellpy/issues/1000
+
+## Original issue text
+
+We do not have any version label for the batch journal file. We should add it. Let us say that if the file misses the version label, it is version 1 (i.e. 1). It can also have the version label with 1 (and new files should be saved with the version number. Then when we decide to change the format, we can bump the version number.
+
+Obviously, we also need to implement reading and saving the version number and prepare for possible version bumps.

cellpy-2.1.4/.issueflows/03-solved-issues/issue1000_plan.md

@@ -0,0 +1,38 @@
+# Issue #1000 — plan: journal JSON version field
+
+## Goal
+
+Batch journal JSON files carry a format `version`. Missing → 1. New writes
+use `JOURNAL_FORMAT_VERSION` (currently 1). Reads store it and warn when the
+file is newer than this cellpy.
+
+## Constraints
+
+- Existing journals without `version` keep loading.
+- Do not implement a v2 schema in this issue — only the hook.
+
+### Prior art
+
+- `JOURNAL_FORMAT_VERSION = 1` already in `cellpy/batch/journal.py` but unused.
+- `read_journal` / `write_journal` — top-level `info_df` / `metadata` / `session`.
+
+## Approach
+
+1. Top-level JSON key `"version"` (integer).
+2. `Journal.version` defaults to 1; `read_journal` fills it (`raw.get("version", 1)`).
+3. `write_journal` always writes `JOURNAL_FORMAT_VERSION`.
+4. If file version > `JOURNAL_FORMAT_VERSION`, `UserWarning` and still load.
+
+## Files to touch
+
+- `cellpy/batch/journal.py`
+- `tests/test_batch_v3.py`
+- `HISTORY.md` (close)
+
+## Test strategy
+
+`uv run pytest tests/test_batch_v3.py` plus `uv run pytest -m essential`.
+
+## Open questions
+
+None.

cellpy-2.1.4/.issueflows/03-solved-issues/issue1000_status.md

@@ -0,0 +1,13 @@
+# Issue #1000 status
+
+- [x] Done
+
+## What's done
+
+- Journal JSON `version` key; missing → 1; writes `JOURNAL_FORMAT_VERSION`.
+- Newer file version warns and still loads.
+- Tests, docs, HISTORY.
+
+## Remaining work
+
+- None.

cellpy-2.1.4/.issueflows/03-solved-issues/issue937_original.md

@@ -0,0 +1,61 @@
+# Issue #937: Notebook tooling (ipykernel, matplotlib) is a hard runtime dependency — ~90 MB in a headless server image
+
+Source: https://github.com/jepegit/cellpy/issues/937
+
+## Original issue text
+
+Found while building a container image for a cellpy-based web app
+([cellpy-simple-gui#121](https://github.com/cellpy/cellpy-simple-gui/issues/121)).
+Not a bug — a packaging pain-point for anyone deploying cellpy headless.
+
+## What we measured
+
+The app plots with plotly, runs FastAPI, and never opens a notebook. Yet the
+image carries an interactive-notebook stack, traced with `importlib.metadata`
+rather than guessed:
+
+```
+debugpy      <- ipykernel
+ipykernel    <- cellpy
+ipython      <- ipykernel
+jedi         <- ipython
+matplotlib   <- cellpy
+```
+
+Sizes in the built image (`python:3.13-slim-bookworm`, 128 packages, 1.18 GB venv):
+
+| package | MB |
+|---|---|
+| `matplotlib` | 35 |
+| `jedi` | 34 |
+| `debugpy` | 22 |
+
+~90 MB, plus `ipython`, `ipykernel`, `pyzmq`, `tornado`, `fontTools` (27 MB) and
+friends behind them. None of it is reachable from the code paths a headless
+server actually uses.
+
+## Why it matters beyond size
+
+- **Container images and frozen apps.** The same weight lands in a PyInstaller
+  bundle, where it is also ~4000 extra files to scan on first run.
+- **Attack surface.** `debugpy` in a server image is not something a deployer
+  would choose.
+- **Cold start.** Not import-time cost (these are lazy), but real disk and pull time.
+
+## Suggestion
+
+Move the interactive pieces to extras and import them where they are used:
+
+```toml
+[project.optional-dependencies]
+notebook = ["ipykernel", "ipython"]
+plotting-mpl = ["matplotlib"]
+```
+
+`cellpy[notebook]` would keep the current experience for notebook users — who
+are surely the majority — while letting an app or a container install the
+analysis core alone. If some module imports `matplotlib` at module scope, a
+local import inside the plotting function would be enough to make the extra
+genuinely optional.
+
+Happy to test a branch against our container build and report the delta.

cellpy-2.1.4/.issueflows/03-solved-issues/issue937_plan.md

@@ -0,0 +1,44 @@
+# Issue #937 plan
+
+## Goal
+
+Drop `ipykernel` and `matplotlib` from the required pip install so a headless
+app image does not pay ~90 MB for notebook/debug tooling it never uses.
+
+## Constraints
+
+- Do not change conda env files (conda-forge still ships the full stack).
+- `uv sync` (dev group) must still install matplotlib so tests/CI keep working.
+- Typed `OptionalDependencyError` naming the extra, same pattern as `legacy-files`.
+- ### Prior art
+  - `require_hdf5_support` in `cellpy/readers/cellpy_file/format.py` + `OptionalDependencyError`
+  - `test_dependency_budget.py` manifest pins for extras moved out of required
+  - `cellpy.plotting.collected` already `try/except ImportError` around matplotlib
+  - `ipykernel` is never imported by library code; only listed in `[project.dependencies]`
+
+## Approach
+
+1. Remove `matplotlib` and `ipykernel` from `[project.dependencies]`.
+2. Add extras `notebook = ["ipykernel", "ipython"]` and `plotting-mpl = ["matplotlib"]`.
+3. Add both packages to the `all` extra; add `matplotlib` to the `dev` group (`ipykernel` is already there).
+4. Add `require_matplotlib(context)` next to the matplotlib backend; call it from `get_backend("matplotlib")`.
+5. Guard the four remaining module-scope matplotlib imports (`figures.py`, `batch_summary.py`, `plotutils.py`, `ocv_rlx.py`) with `try/except ImportError`.
+6. Extend the dependency-budget tests; do not rewrite conda manifests.
+
+## Files to touch
+
+- `pyproject.toml` / `uv.lock` — extras + lock
+- `cellpy/plotting/backends/mpl.py` — `require_matplotlib`
+- `cellpy/plotting/backends/__init__.py` — call it from `get_backend`
+- `cellpy/plotting/figures.py`, `cellpy/plotting/batch_summary.py`, `cellpy/utils/plotutils.py`, `cellpy/utils/ocv_rlx.py` — optional import
+- `tests/test_dependency_budget.py` — pin the new extras
+- `docs/getting_started/agents.md` — one install note
+- `.issueflows/04-designs-and-guides/optional-plotting-notebook.md` — short decision
+
+## Test strategy
+
+`uv run pytest -m essential` plus the new dependency-budget tests. Full suite stays on CI.
+
+## Open questions
+
+None — extras names match the issue (`notebook`, `plotting-mpl`).

cellpy-2.1.4/.issueflows/03-solved-issues/issue937_status.md

@@ -0,0 +1,16 @@
+# Issue #937 status
+
+- [x] Done
+
+## What's done
+
+- Removed `matplotlib` and `ipykernel` from `[project.dependencies]`.
+- Added extras `plotting-mpl` and `notebook`; both included in `all`.
+- `matplotlib` stays in the `dev` group so `uv sync` / CI keep Agg tests.
+- `require_matplotlib` raises `OptionalDependencyError` naming `cellpy[plotting-mpl]`.
+- Guarded the four remaining module-scope matplotlib imports.
+- Dependency-budget tests + docs (`agents.md`, installation extras table).
+
+## Remaining work
+
+- None.

cellpy-2.1.4/.issueflows/03-solved-issues/issue938_original.md

@@ -0,0 +1,73 @@
+# Issue #938: Missing external tools fail silently: mdb-export raises bare FileNotFoundError, pyodbc ImportError hides two loaders
+
+Source: https://github.com/jepegit/cellpy/issues/938
+
+## Original issue text
+
+Two related findings from deploying cellpy in a Linux container
+([cellpy-simple-gui#121](https://github.com/cellpy/cellpy-simple-gui/issues/121)).
+Both are cases where a missing *system* dependency produces a failure that is
+easy to mistake for success.
+
+## 1. `mdb-export` missing → bare `FileNotFoundError`
+
+On posix, `arbin_res` reads `.res` by shelling out to `mdb-export` (mdbtools).
+When it is not installed:
+
+```
+[Errno 2] No such file or directory: 'mdb-export'
+```
+
+An app catching that has to know that "mdb-export" means "install mdbtools", and
+a user reading it in a toast has no chance. It cost us a genuinely misleading
+green tick: our smoke test asserted the import job completed, the job *did*
+complete, and zero cells were imported.
+
+**Suggestion:** raise something named and actionable when the tool is absent,
+e.g.
+
+```
+CellpyDependencyError:
+  Reading Arbin .res on Linux/macOS needs mdbtools (provides `mdb-export`).
+      Debian/Ubuntu:  apt install mdbtools
+      macOS:          brew install mdbtools
+```
+
+A `shutil.which("mdb-export")` check before the call would be enough, and could
+also feed a capability probe (below).
+
+## 2. `libodbc.so.2` missing → two loaders disappear from discovery
+
+`arbin_sql` and `arbin_sql_7` raise at import:
+
+```
+ImportError: libodbc.so.2: cannot open shared object file: No such file or directory
+```
+
+Anything enumerating instruments therefore sees **11 loaders instead of 13**,
+with no indication that two were dropped or why. Our instrument picker simply
+did not offer them, and nothing in the UI could explain the difference between
+the Windows build and the container.
+
+**Suggestion:** let discovery report unavailable loaders rather than omit them —
+something like `(id, available: bool, reason: str)` — so a UI can grey the entry
+out with a tooltip instead of silently narrowing the list. `registry.families()`
+already does something like this well for plots (2.1.2, #868); the same shape
+would suit instruments.
+
+## What this looks like fixed
+
+An app could then ask cellpy "what can you actually read here?" and show the
+answer, instead of discovering the gaps one confused user at a time.
+
+(For the record, the container fix is `apt install mdbtools unixodbc` — but the
+point is that we found it by instrumenting our own test, not from any message
+cellpy produced.)
+
+## Comments (curated summary)
+
+- **Additional tasks**: make the default `paths.examplesdir` absolute (`Path.home() / "cellpy_data" / "examples"`). A relative default resolves against process cwd, so a frozen Windows app wrote demo cells into its install folder.
+- **Clarifications / constraints**: `config.reload()` after setting `CELLPY_PATHS__EXAMPLESDIR` is the workaround that already works; creating the relative directory is not a fix.
+
+_Note: this section is an interpretive summary of the comment thread, not a verbatim dump. Source comments: 1, last comment by @jepegit on 2026-08-16._
+

cellpy-2.1.4/.issueflows/03-solved-issues/issue938_plan.md

@@ -0,0 +1,117 @@
+# Issue #938 plan
+
+Replaces the yolo-abort stub (same filename, in place).
+
+## Goal
+
+Missing *system* tools must be visible: loading an Arbin `.res` without
+`mdb-export` raises a named, installable error; `list_instruments()` still
+lists loaders that failed to import (e.g. `arbin_sql*` without `libodbc.so.2`)
+with `available=False` and a reason a UI can show.
+
+## Constraints
+
+- Stay quiet: `list_instruments` must not emit WARNING spam (#786). Unavailable
+  rows are data, not log noise.
+- Additive listing keys only — existing `id` / `label` / `models` / `suffixes`
+  stay. Apps that ignore unknown keys keep working.
+- Do not change conda env files or add pip extras for mdbtools/unixodbc
+  (system packages).
+- Windows bundled `mdb-export.exe` path stays as-is; the POSIX `PATH` name
+  `"mdb-export"` is the missing-tool case.
+- Public listing shape is documented in `docs/getting_started/agents.md` —
+  update it in the same PR (AGENTS.md “Using cellpy” pointer if the short
+  facts mention the key set).
+- **Out of scope unless Open questions say otherwise:** the comment’s
+  `examplesdir` default. That is cwd/relative-path config, not a missing
+  tool. #960 is a different bug (legacy config file format).
+
+### Prior art
+
+- `OptionalDependencyError` (`cellpy/exceptions.py`) — named missing-dep
+  error used for `tables` / matplotlib; reuse for mdbtools (message names
+  apt/brew, not a pip extra).
+- `arbin_res._loader_posix` already catches `FileNotFoundError`, logs
+  “install mdbtools”, then **re-raises the bare** `FileNotFoundError`
+  (`cellpy/readers/instruments/arbin_res.py`).
+- `cli_api` already probes `command -v mdb-export` on posix (check path
+  only; not used at load time).
+- `InstrumentFactory.create_all(quiet=True)` swallows create failures —
+  that is why `arbin_sql` / `arbin_sql_7` vanish
+  (`cellpy/readers/data_structures.py`).
+- `list_instruments()` (#786) — app picker; tests pin exact key set in
+  `tests/test_instrument_registering.py`.
+- Plot `registry.families()` (#868) lists names that exist; it does **not**
+  actually return `(available, reason)` — mirror the *intent* (UI can enumerate
+  gaps), not the tuple shape.
+- `cellpy.readers.instruments.registry` is the entry-point loader path;
+  built-ins still go through `InstrumentFactory`. Do not mix the two in this
+  issue.
+- Toolbox: no helper for this.
+
+## Approach
+
+**1. Named error when `mdb-export` is missing**
+
+- Small helper next to the loader (e.g. `require_mdb_export(path)`): if `path`
+  is the bare command `mdb-export` (posix) and `shutil.which` is `None`, raise
+  `OptionalDependencyError` naming mdbtools + `apt install mdbtools` /
+  `brew install mdbtools`.
+- If `path` is an absolute/relative file (Windows bundled exe), keep today’s
+  `os.path.isfile` / `FileNotFoundError` behaviour.
+- Call the helper in `_loader_posix` **before** `subprocess.call`, replacing
+  the re-raise of the raw `FileNotFoundError`.
+- Do not lazy-rewrite the whole Arbin loader.
+
+**2. Discovery reports unavailable loaders**
+
+- After `register_builder` for every production loader, iterate **registered
+  ids**, not only `create_all` successes.
+- Success → existing row plus `available: True`, `reason: None`.
+- `create()` / import failure (the `libodbc.so.2` `ImportError`) → still emit
+  a row: `available: False`, `reason: <short exception text>`, empty
+  `models` / `suffixes` if unknown.
+- Expected skips (`local_instrument`, missing `DataLoader`) stay omitted.
+- Optional cheap probe: if `arbin_res` created OK but posix `mdb-export` is
+  missing, mark that row `available: False` with the same mdbtools reason
+  (capability probe the issue asked for). Same helper as part 1.
+- Keep `quiet=True` / DEBUG for the underlying create failures.
+
+**3. Docs / tests**
+
+- `list_instruments` docstring + `agents.md` picker bullet: keys become
+  `{id, label, models, suffixes, available, reason}`.
+- Tests: monkeypatch `shutil.which` → `OptionalDependencyError`; monkeypatch
+  factory create for `arbin_sql` → row present with `available is False`;
+  existing quiet + shape tests updated for the new keys.
+
+## Files to touch
+
+| Path | Change |
+| --- | --- |
+| `cellpy/readers/instruments/arbin_res.py` | `require_mdb_export`; call before subprocess |
+| `cellpy/readers/data_structures.py` | `list_instruments` includes failed creates + `available`/`reason` |
+| `tests/test_instrument_registering.py` | shape keys; unavailable-row test; keep quiet contract |
+| `tests/` (small new or next to arbin tests) | missing `mdb-export` raises named error |
+| `docs/getting_started/agents.md` | picker dict keys |
+| `.issueflows/04-designs-and-guides/instrument-availability.md` | short decision (listing unavailable vs omit) |
+
+## Test strategy
+
+- `uv run pytest -m essential`
+- New/updated tests above; mark essential (public `list_instruments` + load
+  error are app-facing).
+- Full suite on CI.
+
+## Open questions
+
+1. **`examplesdir` comment (frozen-app cwd trap)** — **Recommend: defer.**
+   Different subsystem; #960 is not this bug. Say **include here** if you
+   want a third bullet (`PathsConfig` default + resolve relative →
+   `Path.home() / "cellpy_data" / "examples"` in `example_data`).
+2. **Exception class** — **Recommend: `OptionalDependencyError`.** New
+   `CellpyDependencyError` name from the issue is extra surface for one
+   caller.
+3. **Mark `arbin_res` unavailable when mdbtools is missing?** — **Recommend:
+   yes** (same helper). Listing would otherwise claim the loader works on a
+   container without mdbtools.

cellpy-2.1.4/.issueflows/03-solved-issues/issue938_status.md

@@ -0,0 +1,17 @@
+# Issue #938 status
+
+- [x] Done
+
+## What's done
+
+- `require_mdb_export` / `mdb_export_unavailable_reason` in `arbin_res.py`;
+  posix `.res` load raises `OptionalDependencyError` naming mdbtools.
+- `list_instruments()` walks registered ids; failed creates emit
+  `available=False` + `reason`; expected skips stay omitted; quiet (#786).
+- `arbin_res` listed unavailable when posix `mdb-export` is missing.
+- Tests (essential), `agents.md`, design note `instrument-availability.md`.
+- `examplesdir` deferred (not a missing tool).
+
+## Remaining work
+
+- None.

cellpy-2.1.4/.issueflows/03-solved-issues/issue948_original.md

@@ -0,0 +1,41 @@
+# Issue #948: Problems with custom_group_labels in summary_collector
+
+Source: https://github.com/jepegit/cellpy/issues/948
+
+## Original issue text
+
+Running this:
+```
+cap_summaries = summary_collector(
+    b,
+    #max_cycle=50,
+    columns=[
+        "charge_capacity_gravimetric",
+        "discharge_capacity_gravimetric",
+        "coulombic_efficiency",
+    ],
+    group_it=True,
+    custom_group_labels={
+        1: "run-14",
+        2: "run-15",
+    },
+)
+
+cap_summaries.plot(height=800)
+```
+
+produces expected result with legend labels saying run-14 and run-15, while, if I do:
+
+`cap_summaries.plot(height=800, spread=True)`
+
+changing the the last line to include the standard deviation, the group labels go back to 1 and 2 (The custom_group_labels are still set to run-14 and run-15.) (Showing the standard deviation works)
+
+## Comments (curated summary)
+
+- **Additional tasks**:
+  - Check that `spread=True` still accepts the supported per-panel y-limits (`y_ranges=` / `share_y=False`). The thread's `fig.update_yaxes(..., row=N)` workaround is not the public API and must not become the fix.
+- **Clarifications / constraints**:
+  - Std bands themselves work; only the legend text regresses.
+  - Manual `row=` y-axis edits look wrong on spread because `spread_plot` builds `make_subplots` (row 1 = top) while the workaround assumed the non-spread facet row numbering.
+
+_Note: this section is an interpretive summary of the comment thread, not a verbatim dump. Source comments: 1, last comment by @inger-emma on 2026-09-01._

cellpy-2.1.4/.issueflows/03-solved-issues/issue948_plan.md

@@ -0,0 +1,80 @@
+# Issue #948 plan
+
+## Goal
+
+`summary_collector(..., group_it=True, custom_group_labels={1: "run-14",
+2: "run-15"}).plot(spread=True)` shows **run-14 / run-15** in the Plotly
+legend, same as `.plot()` without spread.
+
+## Constraints
+
+- Additive. Non-spread grouped summaries stay as they are (#923 / #947).
+- Do not invent a second label API. `custom_group_labels=` on collect +
+  `group_label` on the frame is the source of truth.
+- Per-panel y-limits stay `y_ranges=` / `share_y=` (#804 / #817). Do not
+  teach or copy the thread's `fig.update_yaxes(..., row=N)` workaround.
+- Public get/schema/CLI surface unchanged — no agents.md rewrite.
+
+### Prior art
+
+- `cellpy.plotting.collected._spread_series_column` — prefers
+  `group_label` when any value is non-null, else `cell`, else `group`.
+- `spread_plot` — `make_subplots`; `name` / `legendgroup` = groupby key.
+- `summary_plotter` — for `group_it` frames, sets `z` to `group_label`
+  when present (px.line path). Spread does **not** use `z`; it calls
+  `_spread_series_column` on the frame that reaches `sequence_plotter`.
+- `collect_summaries` / `_with_group_label` — writes `group_label` on
+  the averaged long frame from `custom_group_labels` (int or str keys).
+- Tests already cover **Collection**.plot(spread=True) labels
+  (`tests/test_collected_summary_groups.py::test_spread_plot_legend_uses_custom_group_labels`)
+  and **summary_collector**.plot() **without** spread
+  (`test_summary_collector_plot_uses_custom_group_labels_and_units`).
+  The issue snippet — `summary_collector(...).plot(spread=True)` — is
+  **not** locked.
+- Toolbox: none. Graph: none.
+
+## Approach
+
+1. Add the missing essential test: reuse the `summary_collector` batch
+   fixture from `test_summary_collector_plot_uses_custom_group_labels_and_units`,
+   call `.plot(spread=True)`, assert legend names are `run-14` / `run-15`
+   (mean traces with `showlegend`).
+2. Run it. Two outcomes:
+   - **Red** — `group_label` is missing or all-null on the frame
+     `spread_plot` sees (likely dropped in the summary melt / id-vars
+     list, which omits `group_label`). Keep `group_label` as an id
+     column, or ensure `_spread_series_column` sees the same `z` that
+     `summary_plotter` already resolved. Prefer passing the resolved
+     series column into `spread_plot` over a second lookup.
+   - **Green** — already fixed on `master` by #923 / #947; keep the
+     test as the lock and close.
+3. Comment / y-axes: confirm existing
+   `tests/test_collected_summary_axes.py` spread + `y_ranges` tests.
+   If green, one docstring line on `Collection.plot` / `spread_plot`:
+   use `y_ranges=`, not manual `row=`. No start_cell change (spread
+   row 1 = top; tests already assert that order).
+
+## Files to touch
+
+- `tests/test_collected_summary_groups.py` — `summary_collector` +
+  `spread=True` legend test (`@pytest.mark.essential`).
+- `cellpy/plotting/collected.py` — only if the new test is red
+  (`spread_plot` / `_spread_series_column` / melt id-vars).
+- `cellpy/collect/collection.py` — optional one-line `y_ranges=` note
+  on `plot`.
+- `.issueflows/04-designs-and-guides/plotting-collected.md` — spread +
+  `custom_group_labels` / `y_ranges` sentence if code changes.
+
+## Test strategy
+
+```bash
+uv sync --extra batch
+MPLBACKEND=Agg uv run pytest tests/test_collected_summary_groups.py tests/test_collected_summary_axes.py -m essential
+```
+
+## Open questions
+
+- **Comment scope:** treat the y-axis screenshot as “use `y_ranges=`”
+  (recommended) vs also change spread `start_cell` to match the
+  workaround’s row numbers (would invert existing facet-order tests —
+  reject unless you want that).

cellpy-2.1.4/.issueflows/03-solved-issues/issue948_status.md

@@ -0,0 +1,15 @@
+# Issue #948 status
+
+- [x] Done
+
+## What's done
+
+- Plan accepted.
+- `summary_collector(...).plot(spread=True)` legend test — **green** on
+  current master (#923 / #947 already did the plotter work). Test is the lock.
+- `y_ranges=` note on `Collection.plot` / `spread_plot`.
+- HISTORY Unreleased bullet.
+
+## Remaining work
+
+- None.

cellpy-2.1.4/.issueflows/03-solved-issues/issue949_original.md

@@ -0,0 +1,9 @@
+# Issue #949: b.plot not showing ir
+
+Source: https://github.com/jepegit/cellpy/issues/949
+
+## Original issue text
+
+`b.plot(rate=True, ir=True, direction="discharge")`
+
+does not display the IR. The plot looks fine otherwise, and other optional parameters I know of works.