fix(landscape): func_zoning ranks the original P suitability rasters, not C

The reference notebook ("Three scenarios" cells) and zoning_core.py feed E plus
the five P_* WEIGHTED_CLASS rasters into functional zoning; the diagram folder
300-…-Equals_102_202 pointing at the C layers was a drawing error. C_k =
min(P_k, E) stay diagnostic conflict layers. Node 30 keeps its code/stage;
its changed inputs change every fingerprint, so V1–V3 recompute on the next run.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

README.md

@@ -152,7 +152,7 @@ concurrently.
 | 10 | e_ecological_constraint | `lp-z-integrated`, `lp-u-integrated` | `lp-e` |
 | 20 | c_conflicts | `lp-p-irrigation`, `lp-p-snow`, `lp-p-crop`, `lp-p-pasture`, `lp-p-tourism`, `lp-e` | `lp-c-irrigation`, `lp-c-snow`, `lp-c-crop`, `lp-c-pasture`, `lp-c-tourism` |
 | 21 | p_max | five P groups | `lp-p-max` |
-| 30 | func_zoning | `lp-e`, five C groups | `lp-v1-conservative`, `lp-v2-balanced`, `lp-v3-development` |
+| 30 | func_zoning | `lp-e`, five P groups (original suitability) | `lp-v1-conservative`, `lp-v2-balanced`, `lp-v3-development` |
 | 31 | c_max | five C groups | `lp-c-max` |
 
 Sector order everywhere: 1 irrigation, 2 snow, 3 crop, 4 pasture, 5 tourism.
@@ -172,9 +172,10 @@ float32 score to a class: `<= 1.8 → 1`, `<= 2.6 → 2`, `<= 3.4 → 3`, `<= 4.
 | e_ecological_constraint | Z ∧ U valid | `max(Z, U)` |
 | c_conflicts | `P_j` ∧ E valid | `min(P_j, E)` per sector |
 | p_max / c_max | at least one input valid | `max` over the valid ones |
-| func_zoning | E valid ∧ at least one `C_k` valid | winning sector + 1, else 1 |
+| func_zoning | E valid ∧ at least one `P_k` valid | winning sector + 1, else 1 |
 
-Functional zoning takes `S_k = C_k`. A sector is a candidate when `S_k` and `E`
+Functional zoning takes `S_k = P_k`, the original suitability rasters (the conflict
+layers `C_k` are diagnostics and are not zoning inputs). A sector is a candidate when `S_k` and `E`
 are valid and the scenario rule holds — V1: `E <= 2 && S >= v1_min_s_e12` or
 `E == 3 && S >= v1_min_s_e3`; V2: `E <= 2`, or `E == 3 && S >= v2_min_s_e3`, or
 `E == 4 && S >= v2_min_s_e4`; V3: `E <= 4`. With no candidate the pixel is class
@@ -364,6 +365,17 @@ make lint          # golangci-lint (if installed)
 CI (`.github/workflows/ci.yml`) runs build, vet, `go test -race`, and golangci-lint
 on every push and pull request.
 
+Repository tests in `internal/repository/postgres` run against a real database
+when `TEST_DB_URL` points at one with migrations applied, and are skipped
+otherwise. They catch Postgres-only failures (parameter type deduction, casts,
+constraints) that in-memory stubs cannot:
+
+```sh
+docker run -d --name gis-test-pg -e POSTGRES_USER=gis -e POSTGRES_PASSWORD=gis -e POSTGRES_DB=gis -p 5434:5432 postgis/postgis:17-3.5
+DB_URL='postgres://gis:gis@127.0.0.1:5434/gis?sslmode=disable' S3_ENDPOINT=x S3_ACCESS_KEY=x S3_SECRET_KEY=x RABBITMQ_URL=amqp://x go run ./cmd/gis migrate up
+TEST_DB_URL='postgres://gis:gis@127.0.0.1:5434/gis?sslmode=disable' go test ./internal/repository/postgres/ -v
+```
+
 ## Adding a feature
 
 Each new domain is one vertical slice mirroring Category/Dataset:

docs/superpowers/specs/2026-09-05-landscape-program-design.md

@@ -112,11 +112,14 @@ every node of all lower stages finished. Every input and output of every node is
 | 10 | e_ecological_constraint | `lp-z-integrated`, `lp-u-integrated` | `lp-e` |
 | 20 | c_conflicts | `lp-p-irrigation`, `lp-p-snow`, `lp-p-crop`, `lp-p-pasture`, `lp-p-tourism`, `lp-e` | `lp-c-irrigation`, `lp-c-snow`, `lp-c-crop`, `lp-c-pasture`, `lp-c-tourism` |
 | 21 | p_max | five P groups | `lp-p-max` |
-| 30 | func_zoning | `lp-e`, five C groups | `lp-v1-conservative`, `lp-v2-balanced`, `lp-v3-development` |
+| 30 | func_zoning | `lp-e`, five P groups (original suitability) | `lp-v1-conservative`, `lp-v2-balanced`, `lp-v3-development` |
 | 31 | c_max | five C groups | `lp-c-max` |
 
 Sector order everywhere: 1 irrigation, 2 snow, 3 crop, 4 pasture, 5 tourism.
-**Decision:** func_zoning takes S_k = C_k (folder `300-…-Equals_102_202`), not P_k.
+**Decision (revised 2026-09-11):** func_zoning takes S_k = P_k, the original suitability
+rasters, as in the reference notebook ("Three scenarios" cells) and `zoning_core.py`.
+The diagram folder `300-…-Equals_102_202` pointing at the C layers was a drawing error;
+C_k = min(P_k, E) are diagnostic conflict layers and never replace P_k in zoning.
 
 Groups and the category `landscape` ("Ландшафтная программа") are seeded by
 migration `00011_landscape.sql` (`INSERT … ON CONFLICT (code) DO NOTHING`). The
@@ -140,7 +143,7 @@ numpy does with a float32 array.
 | c_conflicts (5 outputs) | P_j ∧ E valid | `min(P_j, E)` |
 | p_max | at least one P valid | `max` over valid P |
 | c_max | at least one C valid | `max` over valid C |
-| func_zoning (3 outputs) | see TZ §4 | see below |
+| func_zoning (3 outputs) | see TZ §4; S_k = P_k | see below |
 
 Functional zoning (TZ `TZ_functional_zoning_go.md`, reference `zoning_core.py`):
 

internal/landscape/graph.go

@@ -89,7 +89,11 @@ var Nodes = []Node{
 	{Code: "10", Name: "e_ecological_constraint", Stage: 1, Inputs: []string{OutZIntegrated, OutUIntegrated}, Outputs: []string{OutE}, Kernel: kernelE},
 	{Code: "20", Name: "c_conflicts", Stage: 2, Inputs: append(append([]string{}, pInputs...), OutE), Outputs: cOutputs, Kernel: kernelCConflicts},
 	{Code: "21", Name: "p_max", Stage: 2, Inputs: pInputs, Outputs: []string{OutPMax}, Kernel: kernelMaxAnyValid},
-	{Code: "30", Name: "func_zoning", Stage: 3, Inputs: append([]string{OutE}, cOutputs...), Outputs: []string{OutV1, OutV2, OutV3}, Kernel: kernelFuncZoning},
+	// func_zoning ranks the original suitability rasters P1..P5 against E (the
+	// reference notebook and zoning_core.py); the conflict layers C_k are
+	// diagnostics and are not zoning inputs. It stays in stage 3 to keep the
+	// diagram's folder numbering (30x) even though it only depends on stage 1.
+	{Code: "30", Name: "func_zoning", Stage: 3, Inputs: append([]string{OutE}, pInputs...), Outputs: []string{OutV1, OutV2, OutV3}, Kernel: kernelFuncZoning},
 	{Code: "31", Name: "c_max", Stage: 3, Inputs: cOutputs, Outputs: []string{OutCMax}, Kernel: kernelMaxAnyValid},
 }
 

internal/landscape/graph_test.go

@@ -132,3 +132,22 @@ func TestNodeForOutput(t *testing.T) {
 		t.Fatal("NodeForOutput(nope) should not be found")
 	}
 }
+
+// Functional zoning ranks the ORIGINAL suitability rasters P1..P5 against E
+// (reference notebook cells "Three scenarios" and zoning_core.py); the
+// conflict layers C_k are diagnostics and must not replace P_k here.
+func TestFuncZoningTakesEAndOriginalP(t *testing.T) {
+	n, ok := NodeByCode("30")
+	if !ok {
+		t.Fatal("node 30 missing")
+	}
+	want := []string{OutE, SlotPIrrigation, SlotPSnow, SlotPCrop, SlotPPasture, SlotPTourism}
+	if len(n.Inputs) != len(want) {
+		t.Fatalf("node 30 inputs = %v, want %v", n.Inputs, want)
+	}
+	for i := range want {
+		if n.Inputs[i] != want[i] {
+			t.Fatalf("node 30 input[%d] = %s, want %s (order is E, then sectors 1..5)", i, n.Inputs[i], want[i])
+		}
+	}
+}

internal/repository/postgres/dataset_db_test.go

@@ -0,0 +1,115 @@
+package postgres
+
+import (
+	"context"
+	"errors"
+	"testing"
+	"time"
+
+	"gis/internal/domain"
+
+	"github.com/google/uuid"
+)
+
+// TestDatasetRepository_GroupsPeriodsAndLatest exercises the Postgres-only
+// parts of the dataset/group repositories added with dataset periods: the
+// ::date casts on *string parameters, the date filter clause, and the
+// "freshest in group" ordering (end_date DESC NULLS LAST, created_at DESC).
+func TestDatasetRepository_GroupsPeriodsAndLatest(t *testing.T) {
+	pool := testPool(t)
+	ctx := context.Background()
+	groups := NewDatasetGroupRepository(pool)
+	datasets := NewDatasetRepository(pool)
+	categories := NewCategoryRepository(pool)
+
+	suffix := uuid.NewString()[:8]
+	cat, err := categories.Create(ctx, domain.Category{Code: "dbtest-" + suffix, Name: "db test"})
+	if err != nil {
+		t.Fatalf("create category: %v", err)
+	}
+	t.Cleanup(func() { _, _ = pool.Exec(ctx, `DELETE FROM categories WHERE id = $1`, cat.ID) })
+
+	group, err := groups.Create(ctx, domain.DatasetGroup{Code: "dbtest-group-" + suffix, Name: "db test group"})
+	if err != nil {
+		t.Fatalf("create group: %v", err)
+	}
+	t.Cleanup(func() { _, _ = pool.Exec(ctx, `DELETE FROM dataset_groups WHERE id = $1`, group.ID) })
+
+	s := func(v string) *string { return &v }
+	mk := func(name string, start, end *string) domain.Dataset {
+		d, err := datasets.Create(ctx, domain.Dataset{
+			CategoryID: cat.ID, DatasetGroupID: &group.ID, Name: name, Filename: name + ".tif",
+			StorageKey: "dbtest/" + suffix + "/" + name, FileType: domain.FileTypeRaster,
+			ContentType: "image/tiff", Status: domain.DatasetStatusReady,
+			StartDate: start, EndDate: end,
+		})
+		if err != nil {
+			t.Fatalf("create dataset %s: %v", name, err)
+		}
+		t.Cleanup(func() { _, _ = pool.Exec(ctx, `DELETE FROM datasets WHERE id = $1`, d.ID) })
+		time.Sleep(5 * time.Millisecond) // distinct created_at for the tie-break
+		return d
+	}
+
+	old := mk("old", s("2019-01-01"), s("2019-12-31"))
+	fresh := mk("fresh", s("2020-01-01"), s("2020-12-31"))
+	undated := mk("undated", nil, nil)
+
+	if old.StartDate == nil || *old.StartDate != "2019-01-01" || *old.EndDate != "2019-12-31" || old.DatasetGroupID == nil {
+		t.Fatalf("dates/group not round-tripped: %+v", old)
+	}
+
+	latest, err := datasets.LatestInGroup(ctx, group.ID, nil, nil)
+	if err != nil {
+		t.Fatalf("LatestInGroup: %v", err)
+	}
+	if latest.ID != fresh.ID {
+		t.Fatalf("latest = %s, want fresh %s (undated %s must sort last)", latest.Name, fresh.ID, undated.ID)
+	}
+	ft := domain.FileTypeVector
+	if _, err := datasets.LatestInGroup(ctx, group.ID, nil, &ft); !errors.Is(err, domain.ErrNotFound) {
+		t.Fatalf("no vector in group: err = %v, want ErrNotFound", err)
+	}
+
+	page, err := datasets.ListSummaries(ctx, domain.DatasetFilter{
+		DatasetGroupID: &group.ID, EndDateFrom: s("2020-01-01"), EndDateTo: s("2020-12-31"),
+	}, 10, 0)
+	if err != nil {
+		t.Fatalf("ListSummaries with date bounds: %v", err)
+	}
+	if len(page) != 1 || page[0].ID != fresh.ID {
+		t.Fatalf("date filter returned %+v, want only fresh", page)
+	}
+	n, err := datasets.Count(ctx, domain.DatasetFilter{DatasetGroupID: &group.ID, StartDateFrom: s("2019-06-01")})
+	if err != nil || n != 1 {
+		t.Fatalf("Count(start_date_from) = %d, %v; want 1", n, err)
+	}
+
+	updated, err := datasets.UpdateFields(ctx, undated.ID, domain.DatasetFields{
+		CategoryID: cat.ID, Name: "undated-now-dated", DatasetGroupID: &group.ID,
+		StartDate: s("2021-01-01"), EndDate: s("2021-12-31"),
+	})
+	if err != nil {
+		t.Fatalf("UpdateFields: %v", err)
+	}
+	if updated.EndDate == nil || *updated.EndDate != "2021-12-31" {
+		t.Fatalf("UpdateFields did not store the period: %+v", updated)
+	}
+	if _, err := datasets.UpdateFields(ctx, undated.ID, domain.DatasetFields{
+		CategoryID: cat.ID, Name: "bad", StartDate: s("2021-12-31"), EndDate: s("2021-01-01"),
+	}); !errors.Is(err, domain.ErrValidation) {
+		t.Fatalf("reversed period must hit the CHECK constraint: err = %v", err)
+	}
+
+	if err := groups.Delete(ctx, group.ID); !errors.Is(err, domain.ErrConflict) {
+		t.Fatalf("deleting a referenced group: err = %v, want ErrConflict", err)
+	}
+	g2, err := groups.SetSLD(ctx, group.ID, "k", "style.sld", "application/xml")
+	if err != nil || g2.SldStorageKey == nil || *g2.SldStorageKey != "k" {
+		t.Fatalf("SetSLD: %+v, %v", g2, err)
+	}
+	byCode, err := groups.GetByCode(ctx, group.Code)
+	if err != nil || byCode.ID != group.ID {
+		t.Fatalf("GetByCode: %+v, %v", byCode, err)
+	}
+}

internal/repository/postgres/landscape.go

@@ -165,9 +165,12 @@ func (r *LandscapeRepository) CountRuns(ctx context.Context) (int, error) {
 // Returns domain.ErrNotFound if the run does not exist.
 func (r *LandscapeRepository) SetRunStatus(ctx context.Context, id uuid.UUID, status domain.LPRunStatus, errMsg *string) error {
 	tag, err := r.pool.Exec(ctx,
-		`UPDATE lp_runs SET status = $2, error = $3,
-		   started_at  = CASE WHEN $2 = 'running' THEN now() ELSE started_at END,
-		   finished_at = CASE WHEN $2 IN ('succeeded','failed') THEN now() ELSE finished_at END
+		// $2 is used both as the assigned value (varchar column) and in the CASE
+		// comparisons (text literals); the explicit ::text casts keep Postgres
+		// from deducing two different types for one parameter (SQLSTATE 42P08).
+		`UPDATE lp_runs SET status = $2::text, error = $3,
+		   started_at  = CASE WHEN $2::text = 'running' THEN now() ELSE started_at END,
+		   finished_at = CASE WHEN $2::text IN ('succeeded','failed') THEN now() ELSE finished_at END
 		 WHERE id = $1`,
 		id, status, errMsg,
 	)

internal/repository/postgres/landscape_db_test.go

@@ -0,0 +1,151 @@
+package postgres
+
+import (
+	"context"
+	"errors"
+	"os"
+	"testing"
+	"time"
+
+	"gis/internal/domain"
+
+	"github.com/google/uuid"
+	"github.com/jackc/pgx/v5/pgxpool"
+)
+
+// testPool connects to the database named by TEST_DB_URL (migrations already
+// applied) or skips the test. Repository SQL is otherwise only exercised by
+// hand, so these tests are the guard against Postgres-only failures such as
+// parameter type deduction (SQLSTATE 42P08) that no stub can reproduce.
+func testPool(t *testing.T) *pgxpool.Pool {
+	t.Helper()
+	url := os.Getenv("TEST_DB_URL")
+	if url == "" {
+		t.Skip("TEST_DB_URL not set; skipping database test")
+	}
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+	pool, err := Connect(ctx, url)
+	if err != nil {
+		t.Fatalf("connect: %v", err)
+	}
+	t.Cleanup(pool.Close)
+	return pool
+}
+
+func TestLandscapeRepository_SetRunStatus_AllTransitions(t *testing.T) {
+	pool := testPool(t)
+	repo := NewLandscapeRepository(pool)
+	ctx := context.Background()
+
+	run, err := repo.CreateRun(ctx, domain.LPRun{
+		Status: domain.LPRunQueued,
+		Params: []byte(`{}`),
+		Nodes:  []domain.LPRunNode{{NodeCode: "00", Stage: 0, Status: domain.LPNodePending}},
+	})
+	if err != nil {
+		t.Fatalf("CreateRun: %v", err)
+	}
+	t.Cleanup(func() { _, _ = pool.Exec(ctx, `DELETE FROM lp_runs WHERE id = $1`, run.ID) })
+
+	if err := repo.SetRunStatus(ctx, run.ID, domain.LPRunRunning, nil); err != nil {
+		t.Fatalf("SetRunStatus(running): %v", err)
+	}
+	got, err := repo.GetRun(ctx, run.ID)
+	if err != nil {
+		t.Fatalf("GetRun: %v", err)
+	}
+	if got.Status != domain.LPRunRunning || got.StartedAt == nil || got.FinishedAt != nil {
+		t.Fatalf("after running: status=%s started=%v finished=%v", got.Status, got.StartedAt, got.FinishedAt)
+	}
+
+	msg := "boom"
+	if err := repo.SetRunStatus(ctx, run.ID, domain.LPRunFailed, &msg); err != nil {
+		t.Fatalf("SetRunStatus(failed): %v", err)
+	}
+	got, err = repo.GetRun(ctx, run.ID)
+	if err != nil {
+		t.Fatalf("GetRun: %v", err)
+	}
+	if got.Status != domain.LPRunFailed || got.FinishedAt == nil || got.Error == nil || *got.Error != msg {
+		t.Fatalf("after failed: status=%s finished=%v error=%v", got.Status, got.FinishedAt, got.Error)
+	}
+
+	if err := repo.SetRunStatus(ctx, run.ID, domain.LPRunSucceeded, nil); err != nil {
+		t.Fatalf("SetRunStatus(succeeded): %v", err)
+	}
+}
+
+func TestLandscapeRepository_NodeLifecycleAndReuse(t *testing.T) {
+	pool := testPool(t)
+	repo := NewLandscapeRepository(pool)
+	ctx := context.Background()
+
+	newRun := func() domain.LPRun {
+		run, err := repo.CreateRun(ctx, domain.LPRun{
+			Status:    domain.LPRunQueued,
+			Params:    []byte(`{"window_size":2048}`),
+			Overrides: map[string]uuid.UUID{"lp-e": uuid.New()},
+			Nodes:     []domain.LPRunNode{{NodeCode: "00", Stage: 0, Status: domain.LPNodePending}},
+		})
+		if err != nil {
+			t.Fatalf("CreateRun: %v", err)
+		}
+		t.Cleanup(func() { _, _ = pool.Exec(ctx, `DELETE FROM lp_runs WHERE id = $1`, run.ID) })
+		return run
+	}
+
+	inputs := []domain.LPDatasetRef{{Group: "lp-z-wr", DatasetID: uuid.New()}}
+	outputs := []domain.LPDatasetRef{{Group: "lp-z-integrated", DatasetID: uuid.New()}}
+
+	first := newRun()
+	if err := repo.SetNodeStarted(ctx, first.ID, "00", "fp-1", inputs); err != nil {
+		t.Fatalf("SetNodeStarted: %v", err)
+	}
+	if err := repo.SetNodeFinished(ctx, first.ID, "00", domain.LPNodeSucceeded, outputs, nil); err != nil {
+		t.Fatalf("SetNodeFinished: %v", err)
+	}
+	got, err := repo.GetRun(ctx, first.ID)
+	if err != nil {
+		t.Fatalf("GetRun: %v", err)
+	}
+	if len(got.Nodes) != 1 || got.Nodes[0].Status != domain.LPNodeSucceeded ||
+		len(got.Nodes[0].Inputs) != 1 || got.Nodes[0].Inputs[0] != inputs[0] ||
+		len(got.Nodes[0].Outputs) != 1 || got.Nodes[0].Outputs[0] != outputs[0] ||
+		got.Nodes[0].StartedAt == nil || got.Nodes[0].FinishedAt == nil {
+		t.Fatalf("node row not round-tripped: %+v", got.Nodes)
+	}
+	if len(got.Overrides) != 1 {
+		t.Fatalf("overrides not round-tripped: %+v", got.Overrides)
+	}
+
+	second := newRun()
+	reused, err := repo.FindReusableOutputs(ctx, "00", "fp-1", second.ID)
+	if err != nil {
+		t.Fatalf("FindReusableOutputs: %v", err)
+	}
+	if len(reused) != 1 || reused[0] != outputs[0] {
+		t.Fatalf("reused = %+v, want %+v", reused, outputs)
+	}
+	if _, err := repo.FindReusableOutputs(ctx, "00", "fp-other", second.ID); !errors.Is(err, domain.ErrNotFound) {
+		t.Fatalf("unknown fingerprint: err = %v, want ErrNotFound", err)
+	}
+	if _, err := repo.FindReusableOutputs(ctx, "00", "fp-1", first.ID); !errors.Is(err, domain.ErrNotFound) {
+		t.Fatalf("own run must be excluded: err = %v, want ErrNotFound", err)
+	}
+
+	failMsg := "grid mismatch"
+	if err := repo.SetNodeFinished(ctx, second.ID, "00", domain.LPNodeFailed, nil, &failMsg); err != nil {
+		t.Fatalf("SetNodeFinished(failed): %v", err)
+	}
+	runs, err := repo.ListRuns(ctx, 10, 0)
+	if err != nil {
+		t.Fatalf("ListRuns: %v", err)
+	}
+	if len(runs) < 2 || runs[0].ID != second.ID {
+		t.Fatalf("ListRuns must be newest first: %+v", runs)
+	}
+	if n, err := repo.CountRuns(ctx); err != nil || n < 2 {
+		t.Fatalf("CountRuns = %d, %v", n, err)
+	}
+}