Merge pull request #87 from ARUP-CAS/version/v2.2.0

Vydání verze 2.2.0: sjednocené popisky filtrů, váha prvku u komponent, ověření přihlášení před stahováním, zapamatované filtry s resetem a aktualizované README.

openspec/changes/archive/2026-10-02-add-filter-memory-and-reset/.openspec.yaml

@@ -0,0 +1,2 @@
+schema: spec-driven
+created: 2026-10-02

openspec/changes/archive/2026-10-02-add-filter-memory-and-reset/design.md

@@ -0,0 +1,172 @@
+# Design
+
+## Context
+
+See proposal.md – Why. Current state of `amcr_viewer/amcr_dialog.py`
+(branch `version/v2.2.0`):
+
+- `AmcrViewer.run_download()` builds a new `AmcrFilterDialog(typ_dat)` for
+  every opening (`amcr_viewer.py`), without a parent.
+- Form state is spread over:
+  - `self.selection_cache` – 19 keys, each a list of codelist codes; the
+    picker's read-only `QLineEdit` text is only set inside the nested
+    `open_dialog()` closure in `setup_picker()`, which keeps no reference
+    to the line edit.
+  - checkboxes `chk_bbox` (default checked), `chk_posevidence`,
+    `chk_proj_akce` (events only), `chk_komponenty` (events and sites);
+  - `self.date_ranges` – `(api_field, name, date_from, date_to)` with
+    nullable `QgsDateEdit`s (`clear()` is the only correct way to empty
+    them, see `_date_edit`).
+- The one non-empty default, *PIAN – přesnost* = `HES-000861/862/863`, is
+  hard-coded inside `setup_picker()` together with its display text.
+- Codelists are module-level dicts in `amcr_codelists.py` mapping
+  **label → code**; `refresh_globals()` updates them in place after
+  *Aktualizovat hesláře*, so the dialog always sees the current values.
+- `tests/smoke_test.py` already builds the dialog offscreen for all three
+  data types and checks `get_filters()` for date ranges.
+
+## Goals / Non-Goals
+
+**Goals:**
+
+- One snapshot format that describes the whole form, used for remember,
+  restore, defaults and "is it default?" comparison.
+- No change in `get_filters()` / `get_bbox()` / `get_komponenty()` output
+  for the same form state.
+
+**Non-Goals:**
+
+- Persisting to `QgsSettings` or the project (variants B and C of #84).
+- Remembering the window size or scroll position.
+
+## Decisions
+
+### Snapshot = plain dict kept at module level in `amcr_dialog.py`
+
+`_REMEMBERED_STATE: dict[str, dict]` keyed by `typ_dat`. The snapshot is
+
+```python
+{
+    "codes": {cache_key: [code, ...], ...},   # only pickers of this typ
+    "checks": {"bbox": bool, "posevidence": bool, ...},
+    "dates": {api_field: (iso_from | None, iso_to | None), ...},
+}
+```
+
+Dates are stored as ISO strings (or `None` for an empty picker), never as
+`QDate`, so a stored value can never turn an empty picker into "today".
+
+- *Why module level, not on `AmcrViewer`:* `run_download()` stays
+  untouched and the smoke test can exercise remember/restore just by
+  creating dialogs. QGIS (and Plugin Reloader) re-imports the plugin
+  package on reload, which drops the dict – that matches the "QGIS run
+  only" requirement.
+- *Alternative – keep one dialog instance per type and only `hide()` it:*
+  rejected. Restoring would be free, but Cancel would then keep the
+  cancelled edits (the requirement says it must not), and long-lived
+  dialogs would hold stale codelist references after an update.
+- *Alternative – pass state in/out through `AmcrViewer`:* works, but adds
+  plumbing in two files for no behavioural gain.
+
+### Defaults defined once
+
+A module-level `DEFAULT_CODES = {"pian_presnost": ["HES-000861",
+"HES-000862", "HES-000863"]}` and `DEFAULT_CHECKS = {"bbox": True}`
+replace the hard-coded block in `setup_picker()`. `_default_state()`
+builds a full snapshot for the dialog's `typ_dat` from them. It is used
+by the constructor (no remembered state), the reset button and the
+"differs from defaults" comparison – one definition, three users.
+
+### Pickers keep a handle to their widgets
+
+`setup_picker()` registers each picker in `self.pickers[cache_key] =
+(data_source, display_field, clear_btn)`. A single
+`_set_picker(cache_key, codes)` sets `selection_cache`, rebuilds the
+display text from the current codelist (inverted `code → label`, sorted
+like the selection dialog), drops unknown codes and enables/disables the
+clear button. `open_dialog()`, restore, reset and clear all go through
+it, so the display text can never disagree with the cache.
+
+Display text is rebuilt from codes rather than stored, so a label renamed
+by a codelist update shows its new name, and a removed code disappears
+(spec: *Restored values follow the current codelists*).
+
+### Remember only in `accept()` after validation
+
+`accept()` already returns early on a reversed date range; the snapshot is
+taken just before `super().accept()`. `reject()` is not overridden.
+
+### Reset button
+
+`QDialogButtonBox.StandardButton.RestoreDefaults` with Czech text
+*Obnovit výchozí* (the standard button would otherwise show the Qt
+translation of "Restore Defaults", which depends on the installed Qt
+translations). Clicking applies `_default_state()` to the form and hides
+the notice; `_REMEMBERED_STATE` is untouched until OK.
+
+The button box already holds *Aktualizovat hesláře* in `ActionRole`; the
+reset button sits next to it on the left, OK/Cancel stay on the right.
+
+### Per-picker clear button
+
+A narrow `QToolButton` with text `✕` next to *Vybrat…*. It calls
+`_set_picker(cache_key, DEFAULT_CODES.get(cache_key, []))` – it returns
+the filter to its default, which is empty for every picker except
+*pian_presnost* (its three pre-selected levels). The tooltip is
+*Vymazat výběr* for an empty default and *Vrátit výchozí výběr* for
+*pian_presnost*. `_set_picker()` enables the button only while the
+current codes differ from the default (compared order-insensitively, so
+a reordered default still counts as the default); on a fresh dialog the
+PIAN button is therefore disabled. For `pian_presnost`, empty means the
+filter is not sent (current `get_filters()` behaviour for an empty
+list) – this matches the spec.
+
+Changed after the user's manual test in QGIS: the ✕ on *PIAN – přesnost*
+emptied the picker, but the user expected it to restore the default
+three levels, so the button now returns each filter to its default
+instead of always emptying it.
+
+Checkboxes and date pickers do not get their own clear button: a
+checkbox is one click, and `QgsDateEdit` with `setAllowNull(True)`
+already has its own clear control.
+
+### Notice about restored filters
+
+A `QLabel` above the bbox checkbox, styled like the existing component
+warning (neutral info colours), hidden by default. Shown in the
+constructor only when a remembered state exists **and** differs from
+`_default_state()`. Text: *Načteny filtry z minulého hledání (aktivní
+filtry: N).* (changed after the user's manual test in QGIS: the leading
+"ℹ " was removed). N counts form items that differ from the default –
+one per picker, checkbox and date row (a date row counts once even
+with both bounds set). Hidden again on reset; not updated live on
+every edit (it describes what was loaded, not the current form).
+
+### Qt5/Qt6
+
+`QToolButton` from `qgis.PyQt.QtWidgets`; all enums fully scoped
+(`QDialogButtonBox.StandardButton.RestoreDefaults`,
+`QDialogButtonBox.ButtonRole.ResetRole`); no `exec_()`.
+
+## Risks / Trade-offs
+
+- [Forgotten filter gives a suspiciously small result] → notice at the top
+  with a count; reset is one click.
+- [Restored bbox restriction with a different map extent] → bbox is a
+  checkbox, the extent itself is read at download time as today; nothing
+  extent-specific is stored.
+- [Codelist update removes a selected code] → dropped silently on
+  restore. Considered warning about it; not done, because the picker text
+  already shows what is selected and the case is rare.
+- [Plugin reload during development keeps the old dict] → only if the
+  package is not re-imported; both QGIS and Plugin Reloader do re-import.
+
+## Verification
+
+- Smoke test (offline, `qgis/qgis:ltr` and `qgis/qgis:stable`): OK →
+  reopen restores codes/checks/dates and `get_filters()` is equal; Cancel
+  keeps the previous state; reset + OK equals a fresh dialog; clear drops
+  one key from `get_filters()`; unknown code is dropped; notice visible
+  only for non-default state; types do not share state.
+- Manual test in QGIS 3.44 and QGIS 4: the scenarios from the spec, plus
+  *Aktualizovat hesláře* between two openings.

openspec/changes/archive/2026-10-02-add-filter-memory-and-reset/proposal.md

@@ -0,0 +1,72 @@
+# Proposal
+
+## Why
+
+The filter dialog (`AmcrFilterDialog`) is created from scratch every time
+the user opens it, so every selection is lost after each download. Refining
+a query ("same area, add one more period") means re-entering every picker
+and date by hand. There is also no quick way back to the default state once
+many filters are set (issue #84).
+
+## What Changes
+
+- The filter dialog remembers the last confirmed filters **per data type**
+  (Fieldwork events, Sites, Individual finds) for the rest of the QGIS
+  run. Reopening the dialog for the same data type restores all pickers,
+  checkboxes and date ranges. Nothing is written to disk; after a QGIS
+  restart (or a plugin reload) the dialog starts from the defaults again.
+- State is remembered only when the dialog is confirmed with OK (after the
+  existing date-range validation passes). *Cancel* leaves the remembered
+  state unchanged.
+- A new *Obnovit výchozí* button (`RestoreDefaults` role) in the button
+  row resets the whole form to its **defaults**, not to an empty form:
+  *Omezit vyhledávání rozsahem okna* checked, *PIAN – přesnost* with its
+  three pre-selected levels, everything else empty. The reset is applied
+  to the form only; the remembered state changes only on OK.
+- When the dialog opens with restored filters that differ from the
+  defaults, a notice at the top says so and how many filters are active,
+  so a forgotten filter further down the scrollable form is not missed.
+- Each picker gets a small clear button (✕) that returns that single
+  filter to its **default** (empty for almost all pickers, the three
+  pre-selected accuracy levels for *PIAN – přesnost*); it is available
+  only while the filter differs from that default. "No PIAN
+  restriction" is still reachable by unchecking all levels in the
+  selection dialog.
+- Restored codes that are no longer in the current codelists (after
+  *Aktualizovat hesláře*) are dropped, and picker texts are rebuilt from
+  the current codelist labels.
+- README (section 3.3) and the v2.2.0 changelog entry in `metadata.txt`
+  describe the new behaviour.
+
+Out of scope:
+
+- Persisting filters across QGIS restarts (`QgsSettings`) or in the QGIS
+  project – considered in issue #84 as variants B and C, not chosen.
+- Sharing filter values between data types.
+
+## Capabilities
+
+### New Capabilities
+
+- `filter-dialog`: state of the filter dialog between openings – remembered
+  filters per data type, reset to defaults, clearing a single filter and
+  the notice about restored filters.
+
+### Modified Capabilities
+
+<!-- none – openspec/specs/ is not maintained (change-tracked) -->
+
+## Impact
+
+- Code: `amcr_viewer/amcr_dialog.py` (state capture/restore, defaults in
+  one place, reset button, per-picker clear button, notice);
+  `tests/smoke_test.py` (offline cases for restore, cancel, reset, clear
+  and dropped codes). `amcr_viewer/amcr_viewer.py` is not expected to
+  change – `run_download` keeps creating the dialog as today.
+- No change to the digiarchiv API requests: `get_filters()`, `get_bbox()`
+  and `get_komponenty()` keep their output for the same form state.
+- No change to layer attributes or stored settings (`QSettings` is not
+  touched).
+- Target branch `version/v2.2.0` (unreleased): the change joins the v2.2.0
+  changelog entry, no separate version bump.
+- Qt5/Qt6 rules from `AGENTS.md` apply; no new dependencies.

openspec/changes/archive/2026-10-02-add-filter-memory-and-reset/specs/filter-dialog/spec.md

@@ -0,0 +1,111 @@
+# Spec Delta
+
+## Purpose
+
+Keeps the filter dialog's selections between openings within one QGIS run,
+so a query can be refined without re-entering it, and gives quick ways back
+to the default state.
+
+## ADDED Requirements
+
+### Requirement: Confirmed filters are remembered per data type
+When the user confirms the filter dialog with OK, the plugin SHALL remember
+the whole form state (all pickers, checkboxes and date ranges) for that
+data type, and SHALL restore it the next time the dialog for the same data
+type is opened within the same QGIS run. Each data type (Fieldwork events,
+Sites, Individual finds) SHALL have its own remembered state.
+
+#### Scenario: Reopening after a download
+- **GIVEN** the user opened the Fieldwork events dialog, selected a region and a period, set a start-date range and confirmed with OK
+- **WHEN** the user opens the Fieldwork events dialog again
+- **THEN** the same region, period and date range are selected and confirming without changes sends the same filters as before
+
+#### Scenario: Data types do not share state
+- **GIVEN** filters were confirmed in the Fieldwork events dialog
+- **WHEN** the user opens the Sites dialog for the first time
+- **THEN** the Sites dialog shows its defaults
+
+#### Scenario: Cancel keeps the previous state
+- **GIVEN** a remembered state exists for a data type
+- **WHEN** the user changes filters and closes the dialog with Cancel
+- **THEN** reopening the dialog shows the remembered state, not the cancelled changes
+
+#### Scenario: Rejected date range is not remembered
+- **WHEN** the user confirms a reversed date range and the dialog refuses it
+- **THEN** the remembered state is unchanged
+
+### Requirement: Remembered state lives only for the QGIS run
+The remembered filters SHALL NOT be written to disk, QGIS settings or the
+project; after QGIS is restarted or the plugin is reloaded, every dialog
+SHALL open with its defaults.
+
+#### Scenario: QGIS restart
+- **GIVEN** filters were confirmed in a previous QGIS run
+- **WHEN** the user opens the dialog after restarting QGIS
+- **THEN** the dialog shows its defaults
+
+### Requirement: Reset restores the defaults
+The filter dialog SHALL offer a reset action that returns every field of
+the form to its default: the map-extent restriction checked, *PIAN –
+přesnost* with its three pre-selected accuracy levels (where the data type
+has it), and every other filter empty. The reset SHALL change only the
+form; the remembered state SHALL change only when the dialog is then
+confirmed with OK.
+
+#### Scenario: Reset and confirm
+- **GIVEN** several filters are set
+- **WHEN** the user resets the form and confirms with OK
+- **THEN** the sent filters equal those of a dialog opened for the first time, and reopening shows the defaults
+
+#### Scenario: Reset and cancel
+- **GIVEN** a remembered state exists
+- **WHEN** the user resets the form and closes the dialog with Cancel
+- **THEN** reopening the dialog shows the remembered state
+
+### Requirement: A single filter can be returned to its default
+Each codelist filter SHALL offer a per-picker action that returns only
+that filter to its default value (empty, or the three pre-selected
+accuracy levels for *PIAN – přesnost*). The action SHALL be available
+only while the filter differs from its default. Returning *PIAN –
+přesnost* to its default SHALL restore the three pre-selected accuracy
+levels; a completely empty *PIAN – přesnost* (no restriction) SHALL
+remain reachable by unchecking all levels in the picker's selection
+dialog.
+
+#### Scenario: Clearing one picker
+- **GIVEN** a region and a period are selected
+- **WHEN** the user clears the region filter
+- **THEN** the region filter shows nothing selected, the period stays selected and the region parameter is not sent
+
+#### Scenario: Returning PIAN to its default
+- **GIVEN** a Fieldwork events dialog is open with *PIAN – přesnost* at its default three accuracy levels
+- **WHEN** the user changes the PIAN selection (for example clears it)
+- **THEN** the picker's clear action becomes available and, when used, restores exactly the three pre-selected accuracy levels
+- **WHEN** the user unchecks all levels in the *PIAN – přesnost* selection dialog instead
+- **THEN** no accuracy restriction is sent
+
+### Requirement: Restored filters are announced
+When the dialog opens with a restored state that differs from the
+defaults, it SHALL show a notice at the top of the form stating that
+filters from the previous search were restored and how many filters
+differ from the defaults. The notice SHALL disappear once the form is
+reset to the defaults.
+
+#### Scenario: Notice after reopening
+- **GIVEN** a region and a period were confirmed
+- **WHEN** the dialog is reopened
+- **THEN** a notice at the top says filters were restored and that 2 filters are active
+
+#### Scenario: No notice for defaults
+- **WHEN** the dialog opens with no remembered state, or with a remembered state equal to the defaults
+- **THEN** no notice is shown
+
+### Requirement: Restored values follow the current codelists
+When restoring, the plugin SHALL drop selected codes that are no longer
+present in the current codelists and SHALL display the remaining
+selections with their current codelist labels.
+
+#### Scenario: Code removed by a codelist update
+- **GIVEN** a confirmed selection contains a code that a later codelist update removed
+- **WHEN** the dialog is reopened
+- **THEN** that code is not selected and not sent, and the other selected values remain

openspec/changes/archive/2026-10-02-add-filter-memory-and-reset/tasks.md

@@ -0,0 +1,82 @@
+# Tasks
+
+## 1. Form state in one place
+
+- [x] 1.1 In `amcr_viewer/amcr_dialog.py` add `DEFAULT_CODES` /
+  `DEFAULT_CHECKS` and `_default_state()`; remove the hard-coded
+  `pian_presnost` block from `setup_picker()` and apply the default
+  through the new path. Verify: smoke test case "filtrační dialogy"
+  still passes and a fresh `akce`/`lokalita` dialog still sends
+  `f_pian_presnost` with the three codes (assert in the new test 1.4)
+- [x] 1.2 Register pickers in `self.pickers` and add `_set_picker()`
+  (cache + display text rebuilt from the current codelist, unknown codes
+  dropped, clear-button state); route `open_dialog()` through it.
+  Verify: `python3 tests/check_sources.py`, `ruff check .`
+- [x] 1.3 Add `_snapshot()` / `_apply_state()` covering codes, checkboxes
+  and date ranges (ISO strings or `None`; empty picker via `clear()`).
+  Verify: smoke test round-trip – snapshot → apply on a fresh dialog →
+  equal `get_filters()`, `get_bbox()`, `get_komponenty()`
+- [x] 1.4 Extend `tests/smoke_test.py` with an offline case for 1.1–1.3
+  (defaults incl. PIAN, round-trip for all three data types, unknown code
+  dropped). Verify: smoke test passes in `qgis/qgis:ltr` and
+  `qgis/qgis:stable`
+
+## 2. Remember, reset, clear, notice
+
+- [x] 2.1 Module-level `_REMEMBERED_STATE` keyed by `typ_dat`; store the
+  snapshot in `accept()` after the date-range check, restore in the
+  constructor. Verify (smoke test): OK → reopen restores; Cancel keeps
+  the previous state; reversed range refused → state unchanged; another
+  data type starts from defaults
+- [x] 2.2 *Obnovit výchozí* button
+  (`QDialogButtonBox.StandardButton.RestoreDefaults`, Czech text) that
+  applies `_default_state()` to the form only. Verify (smoke test): reset
+  + OK equals a fresh dialog; reset + Cancel keeps the remembered state
+- [x] 2.3 `✕` button (`QToolButton`, tooltip *Vymazat výběr*, or
+  *Vrátit výchozí výběr* for a picker with a non-empty default) per
+  picker that returns that filter to its default, enabled only while it
+  differs from the default. Verify (smoke test): clearing one picker
+  removes only its key from `get_filters()`; the PIAN `✕` is disabled
+  on a fresh dialog, enabled after a change and restores the three
+  default levels
+- [x] 2.4 Notice label at the top, shown only when a restored state
+  differs from defaults, with the count of differing items; hidden on
+  reset. Verify (smoke test): hidden for a fresh dialog and for a
+  remembered default state, visible with the right count otherwise
+- [x] 2.5 Reset `_REMEMBERED_STATE` between smoke-test cases (in
+  `try/finally`) so cases stay independent; verify by running the smoke
+  test twice in one container
+
+## 3. Documentation and version
+
+- [x] 3.1 README section 3.3: remembered filters per data type for the
+  QGIS run, *Obnovit výchozí*, `✕` per filter, the notice; adjust the
+  PIAN default note (reset restores it, `✕` clears it). Verify by reading
+  the section against the spec
+- [x] 3.2 Add bullets to the existing v2.2.0 entry of `changelog=` in
+  `amcr_viewer/metadata.txt` (branch `version/v2.2.0` is unreleased, so
+  no new version; `CITATION.cff` already says 2.2.0). Verify:
+  `python3 tests/check_version_bump.py` (or the CI package job) passes
+
+## 4. Final verification
+
+- [x] 4.1 Run the AGENTS.md check set: `tests/check_sources.py`, bandit,
+  detect-secrets `--all-files`, `flake8 --isolated amcr_viewer/`,
+  `ruff check .`, `pyqgis4-checker` (log contains only the header), smoke
+  test in `qgis/qgis:ltr` and `qgis/qgis:stable`; delete
+  `amcr_viewer/__pycache__` afterwards
+- [x] 4.2 `openspec validate add-filter-memory-and-reset --strict` passes
+- [x] 4.3 Manual test in QGIS 3.44 and QGIS 4 (user): spec scenarios –
+  reopen after a download, Cancel, reset + OK / Cancel, `✕` on one
+  picker and on PIAN, notice text, separate state per data type,
+  *Aktualizovat hesláře* between two openings, defaults after a QGIS
+  restart
+  - User: look and function verified on Fieldwork events, Sites and
+    Individual finds; everything worked except two points – the notice
+    started with an odd "ℹ" and `✕` on PIAN emptied it instead of
+    restoring the default. Both fixed (commit d5e520d); the fix was
+    re-tested by the user on Fieldwork events, for Sites and Individual
+    finds it is covered by the smoke test.
+- [x] 4.4 Archive before merge:
+  `openspec archive add-filter-memory-and-reset --skip-specs` in the same
+  PR

openspec/changes/archive/2026-10-02-fix-component-feature-weight/.openspec.yaml

@@ -0,0 +1,2 @@
+schema: spec-driven
+created: 2026-10-02

openspec/changes/archive/2026-10-02-fix-component-feature-weight/design.md

@@ -0,0 +1,49 @@
+## Context
+
+`load_amcr_data` in `amcr_viewer/amcr_tools.py` (section B) builds, for each
+DJ with a PIAN and *Načíst komponenty* on, one metadata dict per component
+and appends it to `pian_lookup[pian_id]`. The weight is set as
+`'vaha': 1/komps_count` with `komps_count = len(komps)` computed **before**
+the loop that skips components failing `komp_projde_filtrem`. The empty-DJ
+branch leaves `vaha` out and the feature builder falls back to
+`meta.get('vaha', 1)`.
+
+## Goals / Non-Goals
+
+**Goals:** weights of one DJ sum to 1 under any filter; the rule is
+testable offline.
+
+**Non-Goals:** weighting across DJs or across records that share one PIAN
+(a PIAN shared by several DJs still yields several features – the weight
+only de-duplicates components of one DJ, as #55 asked); changing the layer
+schema.
+
+## Decisions
+
+1. **Filter first, then weigh.** Build the list of passing components, then
+   set `vaha = 1/len(passing)`. Alternative – a second counting pass with
+   `sum(komp_projde_filtrem(...))` – duplicates the filter call and drifts
+   if the filter changes (#70 replaces it).
+2. **Extract a small pure helper** (e.g. `_component_entries(dj_meta, komps,
+   passes)` returning the list of per-component dicts with `vaha`, where
+   `passes` is a predicate) so the smoke test can check weights without
+   QGIS layers or network. The helper must not depend on how components are
+   selected, so #70 can pass a different predicate.
+3. **Explicit weight 1 for a DJ without components** instead of relying on
+   the `meta.get('vaha', 1)` default – the default stays as a safety net.
+
+## Risks / Trade-offs
+
+- Floating-point: 1/3 weights sum to 0.999…; acceptable for analyses,
+  test with a tolerance.
+- The helper extraction touches a long function; keep the diff limited to
+  the component branch.
+
+## Verification
+
+- Smoke test cases: 4 components no filter → 4×0.25; filter keeps 1 of 4 →
+  weight 1; keeps 2 of 3 → 2×0.5; no components → 1 entry, weight 1.
+- Full AGENTS.md check set (ltr + stable smoke test, pyqgis4-checker).
+- Manual QGIS test by the user: download akce with *Načíst komponenty* and
+  a period filter, check in the attribute table that `prvek_vaha` sums to 1
+  per `dj_id`.

openspec/changes/archive/2026-10-02-fix-component-feature-weight/proposal.md

@@ -0,0 +1,42 @@
+## Why
+
+Issue #55 added the `prvek_vaha` (feature weight) attribute: when *Načíst
+komponenty* is on, every component of a documentation unit (DJ) becomes its
+own feature on the same PIAN geometry, and the weight 1/*n* lets spatial
+analyses count the geometry once. The unreleased implementation on
+`version/v2.2.0` takes *n* from **all** components of the DJ, before the
+period/area filter. With a component filter active the weights of one DJ no
+longer sum to 1 (DJ with 4 components, 1 passes the Neolithic filter → one
+feature with weight 0.25 instead of 1), so weighted counts are wrong exactly
+when users filter. See the comment on #55.
+
+## What Changes
+
+- *n* in `prvek_vaha = 1/n` is the number of component features actually
+  created for the DJ, i.e. components that pass the period/area filters.
+- The weights of all features created from one DJ sum to 1 with or without
+  filters.
+- A DJ without components keeps its single feature with weight 1 (today the
+  value comes from a default; it becomes explicit).
+- `README.md` documents `prvek_vaha` in the component fields table (it is
+  missing there today).
+- Changelog entry under v2.2.0 in `amcr_viewer/metadata.txt` is extended
+  (the feature is unreleased, no separate version bump).
+
+## Capabilities
+
+### New Capabilities
+- `component-features`: one feature per component of a fieldwork event or
+  site, and the weight attribute that de-duplicates shared geometries.
+
+### Modified Capabilities
+
+## Impact
+
+- `amcr_viewer/amcr_tools.py` – component feature creation in
+  `load_amcr_data` (section B, attribute parsing).
+- `tests/smoke_test.py` – offline check of the weights.
+- `README.md`, `amcr_viewer/metadata.txt` (changelog only).
+- No change to the digiarchiv API contract, layer schema or stored settings.
+- `filter-components-via-component-endpoint` (#70) changes how components
+  are selected; it builds on this change and must keep the weight rule.

openspec/changes/archive/2026-10-02-fix-component-feature-weight/specs/component-features/spec.md

@@ -0,0 +1,35 @@
+# Spec Delta
+
+## Purpose
+
+Describes how components of fieldwork events and sites become map features
+and how their weight lets spatial analyses count a shared geometry once.
+
+## ADDED Requirements
+
+### Requirement: Weight of component features sums to one per DJ
+When components are loaded as features, each feature SHALL carry the weight
+`prvek_vaha = 1/n`, where *n* is the number of features created from the
+same documentation unit in this download. Components excluded by the period
+or area filter SHALL NOT count towards *n*.
+
+#### Scenario: No component filter
+- **WHEN** a documentation unit has 4 components and no period or area filter is set
+- **THEN** 4 features are created, each with weight 0.25
+
+#### Scenario: Filter keeps some components
+- **WHEN** a documentation unit has 4 components and the period filter matches 1 of them
+- **THEN** 1 feature is created with weight 1
+
+#### Scenario: Filter keeps two of three components
+- **WHEN** a documentation unit has 3 components and the filter matches 2 of them
+- **THEN** 2 features are created, each with weight 0.5, and their weights sum to 1
+
+### Requirement: Documentation unit without components has weight one
+When components are loaded and a documentation unit has no component, the
+single feature created for it SHALL have weight 1 and empty component
+fields.
+
+#### Scenario: DJ without components, no filter
+- **WHEN** a documentation unit with a PIAN has no components and no component filter is set
+- **THEN** one feature is created with empty component fields and weight 1

openspec/changes/archive/2026-10-02-fix-component-feature-weight/tasks.md

@@ -0,0 +1,37 @@
+# Tasks
+
+## 1. Weight computed from passing components
+
+- [x] 1.1 In `amcr_viewer/amcr_tools.py` extract the per-component entry
+  building of the *Načíst komponenty* branch into a pure helper that takes
+  the DJ metadata, the component list and a pass predicate, filters first
+  and sets `vaha = 1/len(passing)`; set `vaha = 1` explicitly for a DJ
+  without components; verify with `python3 tests/check_sources.py`,
+  `flake8 --isolated amcr_viewer/` and `ruff check .`
+- [x] 1.2 Extend `tests/smoke_test.py` with an offline case for the helper
+  (4 components no filter → 4×0.25; 1 of 4 passes → 1.0; 2 of 3 pass →
+  2×0.5, sum 1 within tolerance; no components → 1 entry, weight 1);
+  verify the smoke test passes in `qgis/qgis:ltr` and `qgis/qgis:stable`
+
+## 2. Documentation
+
+- [x] 2.1 Add `prvek_vaha` (alias *Váha prvku*) to the component fields
+  table in `README.md` with the rule "1/n, n = features created from the
+  same documentation unit after filters"; verify by reading the rendered
+  table
+- [x] 2.2 Extend the v2.2.0 changelog bullet about the feature weight in
+  `amcr_viewer/metadata.txt` (weights of one documentation unit sum to 1
+  also with period/area filters); no version bump – 2.2.0 is unreleased and
+  `CITATION.cff` already says 2.2.0; verify both versions match
+
+## 3. Verification
+
+- [x] 3.1 Run the full local check set from `AGENTS.md` (check_sources,
+  bandit, detect-secrets `--all-files`, flake8 `--isolated`, ruff,
+  pyqgis4-checker log empty, smoke test ltr + stable); verify all clean
+- [x] 3.2 Manual test in QGIS (user): akce in a small window with *Načíst
+  komponenty* and one period filter; verify in the attribute table that
+  `prvek_vaha` sums to 1 per `dj_id`
+  - Verified by the maintainer 2026-10-02: akce and lokality with
+    *Načíst komponenty*, without and with a period filter – weights sum
+    to 1 per DJ and are computed only from the filtered components

openspec/changes/archive/2026-10-02-fix-session-expiry-islogged/.openspec.yaml

@@ -0,0 +1,2 @@
+schema: spec-driven
+created: 2026-10-02

openspec/changes/archive/2026-10-02-fix-session-expiry-islogged/design.md

@@ -0,0 +1,81 @@
+# Design
+
+## Context
+
+- The session lives only in memory (`amcr_tools.AMCR_SESSION`, a
+  `requests.Session` with the `JSESSIONID` cookie). `_get_session()` logs in
+  from stored credentials only when no session object exists, so after QGIS
+  start the first download always logs in; an expired session object is
+  reused forever.
+- All data requests of a download go through `_api_get_json()` (main query
+  pages and PIAN batches). `_is_auth_error()` there reacts to HTTP 401 or
+  error text – neither occurs on expiry (see proposal.md – Why).
+- Server behaviour (verified 2026-10-02, live API):
+  `GET /api/user/islogged` → `{"remaining": <s>}` when logged in,
+  `{"error": "nologged"}` otherwise, both HTTP 200. It does not extend the
+  session; any `search/query` does.
+- Codelists (`amcr_codelists.py`) use plain `requests.get` without the
+  session – unaffected by login state.
+
+## Goals / Non-Goals
+
+**Goals:**
+- One check at the start of `load_amcr_data`, before the first data request.
+- Reuse existing login code (`login_to_api`, `LoginDialog.get_credentials`).
+- Testable offline: the check takes its HTTP behaviour from the session
+  object so the smoke test can inject a fake.
+
+**Non-Goals:**
+- Checking before every page / PIAN batch (a download takes seconds to
+  minutes and every data request renews the sliding timeout).
+- Refactoring session handling into a class.
+
+## Decisions
+
+1. **New helper `_ensure_logged_in() -> str`** in `amcr_tools.py`, returning
+   one of `"anonymous"` (no session, no credentials – nothing to check),
+   `"logged_in"`, `"relogged"`, `"fallback"` (expected login, ended
+   anonymous), `"unknown"` (check failed, proceeding).
+   Flow: get session via `_get_session()` (logs in if needed); if none and no
+   credentials → `anonymous`; if none but credentials → login failed →
+   `fallback`; otherwise call `islogged`; `remaining` → `logged_in`;
+   `nologged` → drop session, re-login once, verify again → `relogged` or
+   `fallback`; exception / non-JSON → `unknown`.
+   *Alternative:* re-login unconditionally before each download – simpler,
+   but one POST with the password per download and no way to distinguish a
+   real failure; rejected.
+   *Alternative:* compare `remaining` with a local timestamp of last request
+   – fragile (server timeout may change); rejected.
+2. **Caller decides UI.** `load_amcr_data` pushes the message bar warning on
+   `fallback`; the helper only logs (keeps it free of `iface` for the test).
+3. **Interpretation of the response:** logged in iff the body is a dict with
+   key `remaining`. Anything else with an `error` key → not logged in. Unknown
+   shape → `unknown` (do not trigger a re-login loop on a format change).
+4. **Keep `_is_auth_error`** as a fallback, with a comment that the current
+   server never triggers it; removing it brings no benefit and it still
+   covers a possible future 401.
+5. **Never log the response of `islogged?wantsUser=true`** – we do not use
+   that parameter at all; only `remaining` (number) is logged.
+
+6. **Logout on credential removal** – new `logout_from_api()` in
+   `amcr_tools.py` called from `LoginDialog._forget_credentials`. The local
+   session is dropped first and unconditionally; the server call is best
+   effort (a failure is logged and reported in the dialog text). Without
+   it, the in-memory session would keep downloading logged-in data until
+   QGIS restarts even though the user believes he is "forgotten".
+
+## Risks / Trade-offs
+
+- [Extra request per download] → only for logged-in / credential users; cost
+  ~100 ms.
+- [Session expires during a very long download] → practically impossible:
+  each page request renews the 1 h sliding timeout.
+- [Re-login prompts for the QGIS master password] → `get_credentials()` is
+  already called on first download after start; behaviour unchanged.
+- [`islogged` endpoint changes shape] → `unknown`, logged warning, download
+  proceeds as today (no regression).
+
+## Migration Plan
+
+Plain plugin update; no settings or data migration. Rollback = previous
+release.

openspec/changes/archive/2026-10-02-fix-session-expiry-islogged/proposal.md

@@ -0,0 +1,59 @@
+# Proposal
+
+## Why
+
+Login to digiarchiv expires after 1 h of inactivity (`sessionTimeout: 3600`)
+and the server then silently treats the request as anonymous: HTTP 200, no
+`error`, only `pristupnost=A` data. The plugin detects expiry only by HTTP 401
+or error text, which never arrives, so a logged-in user who downloads again
+after a pause gets incomplete data without any warning (issue #72, verified
+manually in QGIS and against the live API).
+
+## What Changes
+
+- Before each download the plugin checks the login state with
+  `GET /api/user/islogged` whenever the user is (or should be) logged in –
+  i.e. an in-memory session exists or credentials are stored.
+- When the server answers `{"error": "nologged"}` and credentials are stored,
+  the plugin logs in again and continues the download with the new session.
+- When the re-login fails (or credentials are missing), the plugin warns in
+  the QGIS message bar that the download runs anonymously and returns only
+  records with access level A – not only in the log.
+- Removing the stored credentials (*Odebrat uložené přihlašovací údaje*)
+  also logs the session out on the server (`GET /api/user/logout`) and
+  drops it from memory; today it stays logged in until QGIS restarts.
+- When the check itself cannot be completed (network error, invalid JSON),
+  the download is not blocked; the plugin logs a warning and proceeds.
+- The existing error-text based detection (`_is_auth_error`) stays as
+  a fallback; it is documented as not triggered by the current server.
+- Version bump + changelog (`metadata.txt`, `CITATION.cff`).
+
+Out of scope:
+
+- Keeping the session alive in the background (polling `islogged` does not
+  extend it anyway).
+- Showing the user's access level in the UI (`islogged?wantsUser=true`).
+- Codelist updates: `amcr_codelists` calls the API with plain `requests`
+  without the session, so login state does not affect them today.
+
+## Capabilities
+
+### New Capabilities
+
+- `amcr-session`: login session against digiarchiv – validating the session
+  before a download, transparent re-login and informing the user when data
+  are downloaded anonymously.
+
+### Modified Capabilities
+
+<!-- none – openspec/specs/ is empty -->
+
+## Impact
+
+- Code: `amcr_viewer/amcr_dialog.py` (logout when credentials are
+  removed); `amcr_viewer/amcr_tools.py` (logout helper, new login-state check, call at the start
+  of `load_amcr_data`, message bar warning); `tests/smoke_test.py` (offline
+  test of the check with a mocked HTTP session).
+- API: one extra `GET /api/user/islogged` per download, only when the user is
+  logged in or has stored credentials; anonymous users are unaffected.
+- No new dependencies; Qt5/Qt6 compatibility rules from `AGENTS.md` apply.

openspec/changes/archive/2026-10-02-fix-session-expiry-islogged/specs/amcr-session/spec.md

@@ -0,0 +1,70 @@
+# Spec Delta
+
+## Purpose
+
+Keeps a logged-in user's data downloads from digiarchiv running under a valid
+login, and makes it visible when a download falls back to anonymous access.
+
+## ADDED Requirements
+
+### Requirement: Login state is verified before a download
+Before starting a data download, the plugin SHALL ask the server whether the
+current session is logged in, whenever an in-memory session exists or login
+credentials are stored. Users with neither SHALL download anonymously without
+this check.
+
+#### Scenario: Valid session
+- **WHEN** a session exists and the server reports it as logged in
+- **THEN** the download proceeds with that session and no re-login happens
+
+#### Scenario: Anonymous user without stored credentials
+- **WHEN** no session exists and no credentials are stored
+- **THEN** no login-state request is sent and the download proceeds anonymously without a warning
+
+### Requirement: Expired login is renewed transparently
+When the server reports the session as not logged in and credentials are
+stored, the plugin SHALL log in again and run the whole download with the new
+session.
+
+#### Scenario: Session expired after inactivity
+- **WHEN** the user downloads data more than one hour after the previous download within the same QGIS run
+- **THEN** the plugin logs in again with the stored credentials and the download returns the same records as for a fresh login
+
+#### Scenario: No session yet, credentials stored
+- **WHEN** the first download after QGIS start is requested and credentials are stored
+- **THEN** the plugin logs in and verifies that the new session is logged in before downloading
+
+### Requirement: Anonymous fallback is reported to the user
+When the plugin expected to be logged in but cannot obtain a logged-in
+session, it SHALL show a warning in the QGIS message bar stating that the
+download runs anonymously and contains only records with access level A.
+
+#### Scenario: Re-login fails
+- **WHEN** the session has expired and logging in again with stored credentials fails
+- **THEN** a warning appears in the message bar and the download continues anonymously
+
+#### Scenario: Session expired and credentials removed
+- **WHEN** an in-memory session has expired and no credentials are stored any more
+- **THEN** a warning appears in the message bar and the download continues anonymously
+
+### Requirement: Failed state check does not block the download
+If the login-state check cannot be completed (network error or a response
+that is not valid JSON), the plugin SHALL log a warning and proceed with the
+download using the current session.
+
+#### Scenario: Login-state endpoint unreachable
+- **WHEN** the login-state request fails with a network error
+- **THEN** a warning is written to the log and the download is attempted as usual
+
+### Requirement: Removing stored credentials logs the user out
+When the user removes the stored credentials, the plugin SHALL log the
+current session out on the server and discard it, so that later downloads
+run anonymously without restarting QGIS.
+
+#### Scenario: Credentials removed while logged in
+- **WHEN** the user removes the stored credentials while a logged-in session exists
+- **THEN** the session is logged out on the server and the next download is anonymous without a warning
+
+#### Scenario: Server unreachable during logout
+- **WHEN** the logout request fails with a network error
+- **THEN** the session is still discarded locally and the user is told the next download will be anonymous

openspec/changes/archive/2026-10-02-fix-session-expiry-islogged/tasks.md

@@ -0,0 +1,53 @@
+# Tasks
+
+## 1. Login-state check
+
+- [x] 1.1 Add `_ensure_logged_in()` to `amcr_viewer/amcr_tools.py` per
+  design.md (statuses `anonymous` / `logged_in` / `relogged` / `fallback` /
+  `unknown`, `GET /api/user/islogged` with the current session, one re-login
+  on `nologged`); verify with `python3 tests/check_sources.py` and
+  `ruff check .`
+- [x] 1.2 Add a comment to `_is_auth_error` that the current server never
+  returns such an error on expiry and the check is kept as a fallback;
+  verify by reading the diff
+- [x] 1.3 Extend `tests/smoke_test.py` with offline cases using a fake
+  session object (valid session, `nologged` + successful re-login,
+  `nologged` + failed re-login, no credentials, network error); verify the
+  smoke test passes in `qgis/qgis:ltr` and `qgis/qgis:stable`
+
+## 2. Integration into the download
+
+- [x] 2.1 Call `_ensure_logged_in()` in `load_amcr_data` after the
+  re-entrancy guard, before the first query; on `fallback` push a message
+  bar warning (Czech, scoped `Qgis.MessageLevel.Warning`) that the download
+  runs anonymously and contains only access level A; verify by smoke test
+  and code review
+- [x] 2.2 Live check without credentials: anonymous download path sends no
+  `islogged` request and a made-up `JSESSIONID` yields `nologged`
+  (curl / probe script in scratch); verify outputs recorded in the PR
+- [x] 2.3 Update `README.md` if it describes login/session behaviour; verify
+  the text matches the new behaviour (or note that nothing needed changing)
+
+## 2b. Logout when credentials are removed
+
+- [x] 2b.1 Add `logout_from_api()` to `amcr_tools.py` and call it from
+  `LoginDialog._forget_credentials`; extend the smoke test (session
+  logged out + dropped, network error still drops it, no session = no
+  request); update README and changelog; verify smoke test ltr + stable
+- [x] 2b.2 Manual test in QGIS: log in, download, remove the stored
+  credentials, download again; verify the log shows "Uživatel odhlášen"
+  and the count drops to the anonymous one
+
+## 3. Release preparation and verification
+
+- [x] 3.1 Add changelog entries under v2.2.0 in `amcr_viewer/metadata.txt`
+  (the fix ships with 2.2.0; `CITATION.cff` already says 2.2.0 and
+  `date-released` moves on release day); verify both versions match
+- [x] 3.2 Run the full local check set from `AGENTS.md` (check_sources,
+  bandit, detect-secrets `--all-files`, flake8 `--isolated`, ruff,
+  pyqgis4-checker log empty, smoke test ltr + stable); verify all clean
+- [x] 3.3 Manual test in QGIS with a researcher account: download SN for
+  whole CZ, simulate expiry in the Python console with
+  `amcr_tools.AMCR_SESSION.get("https://digiarchiv.aiscr.cz/api/user/logout")`,
+  download again; verify log shows re-login and the count matches the
+  logged-in count (not the anonymous one)