sm-bm-plugin/controllers/api/material_search_test.go

package api_test

import (
	"encoding/json"
	"net/http"
	"net/http/httptest"
	"strings"
	"sync"
	"testing"

	"git.golem15.com/golem15/summercms/modules/beachcomber"
	"git.golem15.com/golem15/summercms/modules/bouncer"
	"git.golem15.com/jakub/sm-bm-plugin/classes"
	"git.golem15.com/jakub/sm-bm-plugin/controllers/api"
	"git.golem15.com/jakub/sm-bm-plugin/internal/searchtest"
	"git.golem15.com/jakub/sm-bm-plugin/models"
	"gorm.io/driver/postgres"
	"gorm.io/gorm"
)

func segments(parts ...any) []beachcomber.Segment {
	out := make([]beachcomber.Segment, 0, len(parts)/2)
	for i := 0; i+1 < len(parts); i += 2 {
		out = append(out, beachcomber.Segment{Text: parts[i].(string), Match: parts[i+1].(bool)})
	}
	return out
}

func TestMaterialSnippet(t *testing.T) {
	body := "# Wstęp\n\nNa początku. Proces **globalizacji** zmienia państwa. Koniec tekstu."
	m := models.Material{ID: 3, Body: &body}
	hit := beachcomber.Hit{ID: "3", Highlights: []beachcomber.Snippet{
		{Field: "title", ArrayIndex: -1, Segments: segments("Tytuł", true)},
		{Field: "body", ArrayIndex: -1, Segments: segments("Proces ", false, "globalizacji", true, " zmienia <b>państwa</b>.", false)},
	}}
	s := api.PickSnippet(hit, m)
	if s == nil {
		t.Fatal("no snippet")
	}
	if s.Source != "body" || s.FileName != nil {
		t.Errorf("source %q file %v, want body and no file", s.Source, s.FileName)
	}
	if len(s.Segments) != 3 || s.Segments[1].Text != "globalizacji" || !s.Segments[1].Hit || s.Segments[0].Hit {
		t.Errorf("segments = %+v", s.Segments)
	}
	if s.Segments[2].Text != " zmienia <b>państwa</b>." {
		t.Errorf("segment text was changed: %q", s.Segments[2].Text)
	}
	if !s.Lead || !s.Trail {
		t.Errorf("lead %v trail %v, want both (cut inside the body)", s.Lead, s.Trail)
	}
	b, err := json.Marshal(s)
	if err != nil {
		t.Fatal(err)
	}
	if got := string(b); !strings.Contains(got, `"source":"body","file_name":null,"lead":true,"trail":true,"segments":[{"text":"Proces ","hit":false}`) {
		t.Errorf("JSON = %s", got)
	}

	// The whole body: neither lead nor trail.
	short := "Krótka **notatka**"
	whole := api.PickSnippet(beachcomber.Hit{Highlights: []beachcomber.Snippet{{Field: "body", ArrayIndex: -1, Segments: segments("Krótka ", false, "notatka", true)}}}, models.Material{Body: &short})
	if whole == nil || whole.Lead || whole.Trail {
		t.Errorf("whole body snippet = %+v, want no lead or trail", whole)
	}
	// Start of the body only: trail but no lead.
	start := api.PickSnippet(beachcomber.Hit{Highlights: []beachcomber.Snippet{{Field: "body", ArrayIndex: -1, Segments: segments("Wstęp ", false, "Na", true)}}}, m)
	if start == nil || start.Lead || !start.Trail {
		t.Errorf("start snippet = %+v, want trail only", start)
	}

	// A snippet on another field only, or none at all, gives no snippet.
	if got := api.PickSnippet(beachcomber.Hit{Highlights: []beachcomber.Snippet{{Field: "title", ArrayIndex: -1, Segments: segments("x", true)}}}, m); got != nil {
		t.Errorf("title-only hit gave %+v", got)
	}
	if got := api.PickSnippet(beachcomber.Hit{ID: "3"}, m); got != nil {
		t.Errorf("hit without highlights gave %+v", got)
	}

	if got := api.FlattenSnippet(s); got != "…Proces globalizacji zmienia <b>państwa</b>.…" {
		t.Errorf("FlattenSnippet = %q", got)
	}
	if got := api.FlattenSnippet(nil); got != "" {
		t.Errorf("FlattenSnippet(nil) = %q", got)
	}
	if got := api.RankOrder([]uint{9, 3}); got != "array_position(ARRAY[9,3]::bigint[], jz_bm_materials.id), jz_bm_materials.id DESC" {
		t.Errorf("RankOrder = %q", got)
	}
}

func TestMaterialsSearchUnavailable(t *testing.T) {
	h := noDB()
	user := &bouncer.Principal{ID: 7}
	code, body := serve(t, h.Materials, http.MethodGet, "/?q=globalizacja", user, nil)
	if code != http.StatusServiceUnavailable || body["error"] != true || body["message"] != "Search is unavailable." {
		t.Errorf("q without an index = %d %v, want 503 Search is unavailable.", code, body)
	}
	// One rune is no search: the list answers as without q (no database).
	code, body = serve(t, h.Materials, http.MethodGet, "/?q=%20g%20", user, nil)
	if code != http.StatusServiceUnavailable || body["message"] != "Database unavailable." {
		t.Errorf("one-rune q = %d %v, want the no-database answer", code, body)
	}
	code, body = serve(t, h.Materials, http.MethodGet, "/", user, nil)
	if code != http.StatusServiceUnavailable || body["message"] != "Database unavailable." {
		t.Errorf("no q = %d %v, want the no-database answer", code, body)
	}
	// An unconfigured engine is the same 503.
	h.MaterialIndex = classes.NewMaterialIndexer(searchtest.App(t))
	code, body = serve(t, h.Materials, http.MethodGet, "/?q=globalizacja", user, nil)
	if code != http.StatusServiceUnavailable || body["message"] != "Search is unavailable." {
		t.Errorf("q with a null engine = %d %v, want 503 Search is unavailable.", code, body)
	}
}

// sqlRecorder is a dry-run postgres handle that records every query it
// builds, with its arguments inlined.
type sqlRecorder struct {
	mu  sync.Mutex
	sql []string
}

func (r *sqlRecorder) queries() []string {
	r.mu.Lock()
	defer r.mu.Unlock()
	return append([]string(nil), r.sql...)
}

func dryDB(t *testing.T) (*gorm.DB, *sqlRecorder) {
	t.Helper()
	db, err := gorm.Open(postgres.New(postgres.Config{DSN: "host=127.0.0.1"}), &gorm.Config{DryRun: true, DisableAutomaticPing: true})
	if err != nil {
		t.Fatal(err)
	}
	rec := &sqlRecorder{}
	err = db.Callback().Query().After("gorm:query").Register("test:record_sql", func(tx *gorm.DB) {
		rec.mu.Lock()
		defer rec.mu.Unlock()
		rec.sql = append(rec.sql, tx.Dialector.Explain(tx.Statement.SQL.String(), tx.Statement.Vars...))
	})
	if err != nil {
		t.Fatal(err)
	}
	return db, rec
}

func TestMaterialsSearchTrace(t *testing.T) {
	stub := &searchtest.Stub{Hits: []searchtest.Hit{
		{ID: "9"},
		{ID: "3", Field: "body", Snippet: "o " + searchtest.HighlightStart + "globalizacji" + searchtest.HighlightEnd},
	}}
	db, rec := dryDB(t)
	h := &api.Handlers{
		DB:            func() *gorm.DB { return db },
		MaterialIndex: classes.NewMaterialIndexer(searchtest.App(t, searchtest.Start(t, stub)...)),
	}
	req := httptest.NewRequest(http.MethodGet, "/materials?q=globalizacja&course=7", nil)
	req = req.WithContext(bouncer.WithUser(req.Context(), &bouncer.Principal{ID: 5}))
	w := httptest.NewRecorder()
	h.Materials(w, req)
	if w.Code != http.StatusOK {
		t.Fatalf("status %d: %s", w.Code, w.Body.String())
	}
	if n := len(stub.Searches()); n != 1 {
		t.Errorf("searches = %d, want 1", n)
	}
	var count, page string
	for _, q := range rec.queries() {
		switch {
		case strings.HasPrefix(q, "SELECT count(*)"):
			count = q
		case strings.HasPrefix(q, "SELECT") && strings.Contains(q, `FROM "jz_bm_materials"`):
			page = q
		}
	}
	if count == "" || page == "" {
		t.Fatalf("queries = %q, want a count and a page", rec.queries())
	}
	for name, q := range map[string]string{"count": count, "page": page} {
		for _, want := range []string{"jz_bm_materials.is_private = FALSE", "jz_bm_materials.published_at IS NOT NULL", "jz_bm_materials.course_id = 7", "jz_bm_materials.id IN (9,3)"} {
			if !strings.Contains(q, want) {
				t.Errorf("%s query lacks %q:\n%s", name, want, q)
			}
		}
	}
	if !strings.Contains(page, "array_position(ARRAY[9,3]::bigint[], jz_bm_materials.id)") {
		t.Errorf("page query is not in rank order:\n%s", page)
	}
	var out struct {
		Data []map[string]any `json:"data"`
		Meta struct {
			Total int `json:"total"`
		} `json:"meta"`
	}
	if err := json.Unmarshal(w.Body.Bytes(), &out); err != nil {
		t.Fatal(err)
	}
	if out.Meta.Total != 0 || len(out.Data) != 0 {
		t.Errorf("total %d data %v, want the dry SQL's 0 (never the engine's found)", out.Meta.Total, out.Data)
	}
}

func TestManagedMaterialsSearch(t *testing.T) {
	h := noDB()
	user := &bouncer.Principal{ID: 7}
	code, body := serve(t, h.ManagedMaterials, http.MethodGet, "/?q=globalizacja", user, nil)
	if code != http.StatusServiceUnavailable || body["error"] != true || body["message"] != "Search is unavailable." {
		t.Errorf("q without an index = %d %v, want 503 Search is unavailable.", code, body)
	}
	code, body = serve(t, h.ManagedMaterials, http.MethodGet, "/?q="+strings.Repeat("a", 101), user, nil)
	if code != http.StatusUnprocessableEntity {
		t.Errorf("q over 100 characters = %d %v, want 422", code, body)
	} else if errs, _ := body["errors"].(map[string]any); errs["q"] == nil {
		t.Errorf("422 errors %v have no q message", errs)
	}
	code, body = serve(t, h.ManagedMaterials, http.MethodGet, "/?q=a&include_body=true", user, nil)
	if code != http.StatusServiceUnavailable || body["message"] != "Database unavailable." {
		t.Errorf("one-rune q = %d %v, want the no-database answer", code, body)
	}
}

func TestMaterialSnippetMCP(t *testing.T) {
	body := "Wstęp. Proces globalizacji zmienia państwa. Koniec."
	m := models.Material{Body: &body}
	mid := api.PickSnippet(beachcomber.Hit{Highlights: []beachcomber.Snippet{{Field: "body", ArrayIndex: -1, Segments: segments("Proces ", false, "globalizacji", true, " zmienia", false)}}}, m)
	if got := api.FlattenSnippet(mid); got != "…Proces globalizacji zmienia…" {
		t.Errorf("middle = %q, want ellipses on both sides", got)
	}
	head := api.PickSnippet(beachcomber.Hit{Highlights: []beachcomber.Snippet{{Field: "body", ArrayIndex: -1, Segments: segments("Wstęp", true, ". Proces", false)}}}, m)
	if got := api.FlattenSnippet(head); got != "Wstęp. Proces…" {
		t.Errorf("head = %q, want a trailing ellipsis only", got)
	}
	tail := api.PickSnippet(beachcomber.Hit{Highlights: []beachcomber.Snippet{{Field: "body", ArrayIndex: -1, Segments: segments("państwa. ", false, "Koniec", true, ".", false)}}}, m)
	if got := api.FlattenSnippet(tail); got != "…państwa. Koniec." {
		t.Errorf("tail = %q, want a leading ellipsis only", got)
	}
	whole := api.PickSnippet(beachcomber.Hit{Highlights: []beachcomber.Snippet{{Field: "body", ArrayIndex: -1, Segments: segments(body, false)}}}, m)
	if got := api.FlattenSnippet(whole); got != body {
		t.Errorf("whole = %q, want no ellipsis", got)
	}
}