This commit is contained in:
2026-08-16 18:36:56 +02:00
parent d981147ff2
commit a94a5aaad7
26 changed files with 904 additions and 507 deletions
+19 -196
View File
@@ -1,26 +1,15 @@
package main
import (
"bufio"
"context"
"fmt"
"io"
"log"
"os"
"strings"
"github.com/descrybe/descrybe-v2/apps/api/internal/billing"
"github.com/descrybe/descrybe-v2/apps/api/internal/processing"
"github.com/google/uuid"
"github.com/jackc/pgx/v5/pgxpool"
)
// Classic A1 Postman / Elkotex fixture EANs — must never live on Platform Demo.
var a1FixtureEANs = []string{
"5905575903198",
"6970995789942",
}
type categoryBackfillResult struct {
ProcessedUpdated int64
ProcessedInserted int64
@@ -34,204 +23,38 @@ type categoryBackfillResult struct {
MappedWithoutCat int
}
// backfillMappedCategoriesFromProcessed copies processed_products.category into
// raw_products.mapped_data.category for A1 only. Delegates to processing.
// a1FixtureEANs aliases processing.A1FixtureEANs for seed-a1 tests.
var a1FixtureEANs = processing.A1FixtureEANs
func backfillMappedCategoriesFromProcessed(ctx context.Context, pg *pgxpool.Pool, companyID uuid.UUID) (int64, error) {
return processing.BackfillMappedCategoriesFromProcessed(ctx, pg, companyID)
}
// backfillCategoriesFromMySQLDump streams dump processed_products for the A1
// legacy company and writes product_id (GTIN) → category onto A1 Postgres
// mapped_data.category (and updates any existing processed_products.category).
// It does not insert processed rows — A1 demo seed stays at processed=0.
func backfillCategoriesFromMySQLDump(ctx context.Context, pg *pgxpool.Pool, companyID uuid.UUID, dumpPath string) (categoryBackfillResult, error) {
var out categoryBackfillResult
legacyCompany := billing.A1LegacyCompanyID
var legacy string
_ = pg.QueryRow(ctx, `SELECT COALESCE(legacy_company_id::text, '') FROM companies WHERE id = $1`, companyID).Scan(&legacy)
if legacy != "" {
legacyCompany = legacy
}
f, err := os.Open(dumpPath)
res, err := processing.BackfillCategoriesFromMySQLDump(ctx, pg, companyID, dumpPath)
if err != nil {
return out, fmt.Errorf("open mysql dump: %w", err)
return categoryBackfillResult{}, err
}
defer f.Close()
byGTIN, err := scanA1ProcessedCategories(f, legacyCompany)
if err != nil {
return out, err
}
out.DumpPairs = len(byGTIN)
if len(byGTIN) == 0 {
return out, fmt.Errorf("no A1 processed_products categories for legacy %s in dump", legacyCompany)
}
log.Printf("dump: %d A1 gtin→category pairs", len(byGTIN))
gtins := make([]string, 0, len(byGTIN))
cats := make([]string, 0, len(byGTIN))
for g, c := range byGTIN {
gtins = append(gtins, g)
cats = append(cats, c)
}
ct, err := pg.Exec(ctx, `
UPDATE processed_products p
SET category = v.category,
updated_at = now()
FROM unnest($2::text[], $3::text[]) AS v(gtin, category)
WHERE p.company_id = $1
AND p.product_id = v.gtin
AND COALESCE(NULLIF(trim(v.category), ''), '') <> ''
AND (
COALESCE(NULLIF(trim(p.category), ''), '') = ''
OR lower(trim(p.category)) = 'none'
OR p.category IS DISTINCT FROM v.category
)`, companyID, gtins, cats)
if err != nil {
return out, fmt.Errorf("update processed category from dump: %w", err)
}
out.ProcessedUpdated = ct.RowsAffected()
// A1 demo seed keeps processed=0. Do not INSERT processed rows from the dump —
// only refresh mapped_data.category (and any existing processed rows if present).
ct, err = pg.Exec(ctx, `
UPDATE raw_products r
SET mapped_data = jsonb_set(
COALESCE(r.mapped_data, '{}'::jsonb),
'{category}',
to_jsonb(v.category),
true
),
updated_at = now()
FROM unnest($2::text[], $3::text[]) AS v(gtin, category)
WHERE r.company_id = $1
AND r.gtin = v.gtin
AND COALESCE(NULLIF(trim(v.category), ''), '') <> ''
AND (
COALESCE(NULLIF(trim(r.mapped_data->>'category'), ''), '') = ''
OR r.mapped_data->>'category' IS DISTINCT FROM v.category
)`, companyID, gtins, cats)
if err != nil {
return out, fmt.Errorf("update mapped category from dump: %w", err)
}
out.MappedUpdated = ct.RowsAffected()
n, err := backfillMappedCategoriesFromProcessed(ctx, pg, companyID)
if err != nil {
return out, err
}
out.MappedUpdated += n
if err := fillCategoryCoverage(ctx, pg, companyID, &out); err != nil {
return out, err
}
return out, nil
return categoryBackfillResult{
ProcessedUpdated: res.ProcessedUpdated,
ProcessedInserted: res.ProcessedInserted,
MappedUpdated: res.MappedUpdated,
DumpPairs: res.DumpPairs,
PurgedOtherRaw: res.PurgedOtherRaw,
PurgedOtherPP: res.PurgedOtherPP,
ProcessedWithCat: res.ProcessedWithCat,
ProcessedWithoutCat: res.ProcessedWithoutCat,
MappedWithCat: res.MappedWithCat,
MappedWithoutCat: res.MappedWithoutCat,
}, nil
}
func fillCategoryCoverage(ctx context.Context, pg *pgxpool.Pool, companyID uuid.UUID, out *categoryBackfillResult) error {
err := pg.QueryRow(ctx, `
SELECT
COUNT(*) FILTER (
WHERE COALESCE(NULLIF(trim(category), ''), '') <> ''
AND lower(trim(category)) <> 'none'
),
COUNT(*) FILTER (
WHERE COALESCE(NULLIF(trim(category), ''), '') = ''
OR lower(trim(category)) = 'none'
)
FROM processed_products
WHERE company_id = $1`, companyID).Scan(&out.ProcessedWithCat, &out.ProcessedWithoutCat)
if err != nil {
return fmt.Errorf("count processed categories: %w", err)
}
err = pg.QueryRow(ctx, `
SELECT
COUNT(*) FILTER (
WHERE COALESCE(NULLIF(trim(mapped_data->>'category'), ''), '') <> ''
),
COUNT(*) FILTER (
WHERE COALESCE(NULLIF(trim(mapped_data->>'category'), ''), '') = ''
)
FROM raw_products
WHERE company_id = $1`, companyID).Scan(&out.MappedWithCat, &out.MappedWithoutCat)
if err != nil {
return fmt.Errorf("count mapped categories: %w", err)
}
return nil
}
// purgeA1FixtureEANsFromOtherTenants deletes the Postman Elkotex fixture EANs
// from every company except A1 (Platform Demo must not mirror A1 fixtures).
func purgeA1FixtureEANsFromOtherTenants(ctx context.Context, pg *pgxpool.Pool, a1CompanyID uuid.UUID) (rawN, ppN int64, err error) {
ct, err := pg.Exec(ctx, `
DELETE FROM processing_job_products pjp
WHERE pjp.raw_product_id IN (
SELECT id FROM raw_products
WHERE company_id <> $1 AND gtin = ANY($2::text[])
)
OR pjp.processed_product_id IN (
SELECT id FROM processed_products
WHERE company_id <> $1 AND product_id = ANY($2::text[])
)`, a1CompanyID, a1FixtureEANs)
if err != nil {
return 0, 0, fmt.Errorf("purge fixture job products: %w", err)
}
_ = ct
ct, err = pg.Exec(ctx, `
DELETE FROM processed_products
WHERE company_id <> $1 AND product_id = ANY($2::text[])`, a1CompanyID, a1FixtureEANs)
if err != nil {
return 0, 0, fmt.Errorf("purge fixture processed: %w", err)
}
ppN = ct.RowsAffected()
ct, err = pg.Exec(ctx, `
DELETE FROM raw_products
WHERE company_id <> $1 AND gtin = ANY($2::text[])`, a1CompanyID, a1FixtureEANs)
if err != nil {
return 0, 0, fmt.Errorf("purge fixture raw: %w", err)
}
rawN = ct.RowsAffected()
return rawN, ppN, nil
return processing.PurgeA1FixtureEANsFromOtherTenants(ctx, pg, a1CompanyID)
}
func scanA1ProcessedCategories(r io.Reader, legacyCompany string) (map[string]string, error) {
br := bufio.NewReaderSize(r, 1<<20)
inTable := false
out := make(map[string]string, 4096)
for {
line, err := br.ReadString('\n')
if len(line) > 0 {
trimmed := strings.TrimSpace(line)
if strings.HasPrefix(trimmed, "INSERT INTO `processed_products`") ||
strings.HasPrefix(trimmed, "INSERT INTO processed_products") {
inTable = true
} else if inTable && strings.HasPrefix(trimmed, "CREATE TABLE") {
break
} else if inTable && strings.HasPrefix(trimmed, "INSERT INTO `") &&
!strings.Contains(trimmed, "processed_products") {
break
} else if inTable && looksLikeTupleLine(line) && strings.Contains(line, legacyCompany) {
fields := parseMySQLTupleFieldsN(line, 6)
if len(fields) >= 5 {
gtin := strings.TrimSpace(fields[2])
cat := strings.TrimSpace(fields[4])
if gtin != "" && cat != "" && !strings.EqualFold(cat, "NULL") {
out[gtin] = cat
}
}
}
}
if err == io.EOF {
break
}
if err != nil {
return nil, err
}
}
return out, nil
return processing.ScanA1ProcessedCategories(r, legacyCompany)
}
func logCategoryBackfillResult(res categoryBackfillResult, source string) {
+10 -52
View File
@@ -4,70 +4,28 @@ import (
"context"
"fmt"
"log"
"os"
"path/filepath"
"strings"
"github.com/descrybe/descrybe-v2/apps/api/internal/processing"
"github.com/google/uuid"
"github.com/jackc/pgx/v5/pgxpool"
)
// resolveMySQLDumpPath picks an explicit path, else the first readable candidate
// under common local locations documented in scripts/seed/README.txt.
// resolveMySQLDumpPath delegates to processing (shared with admin Sync A1).
func resolveMySQLDumpPath(explicit string) string {
if p := strings.TrimSpace(explicit); p != "" {
if st, err := os.Stat(p); err == nil && !st.IsDir() {
return p
}
log.Printf("warning: mysql dump not found at %q — trying auto-detect", p)
}
for _, c := range mysqlDumpCandidates() {
if strings.TrimSpace(explicit) != "" && filepath.Clean(c) == filepath.Clean(strings.TrimSpace(explicit)) {
continue
}
if st, err := os.Stat(c); err == nil && !st.IsDir() {
return c
}
}
return ""
return processing.ResolveMySQLDumpPath(explicit)
}
func mysqlDumpCandidates() []string {
var out []string
if v := strings.TrimSpace(os.Getenv("SEED_A1_MYSQL_DUMP")); v != "" {
out = append(out, v)
}
home, _ := os.UserHomeDir()
names := []string{
"descrybe_new (1).sql",
"descrybe_new.sql",
"descrybe_new(1).sql",
}
if home != "" {
for _, n := range names {
out = append(out, filepath.Join(home, "Downloads", n))
out = append(out, filepath.Join(home, "downloads", n))
}
}
// Repo-relative guesses (cwd may be apps/api or repo root).
for _, n := range names {
out = append(out,
n,
filepath.Join("..", "..", n),
filepath.Join("scripts", "seed", n),
filepath.Join("..", "..", "scripts", "seed", n),
)
}
return out
return processing.MySQLDumpCandidates()
}
type mappedCoverage struct {
Total int
WithDesc int
WithCat int
WithAttrs int
Processed int
Jobs int
Total int
WithDesc int
WithCat int
WithAttrs int
Processed int
Jobs int
}
func (c mappedCoverage) pct(n int) float64 {
+6 -2
View File
@@ -41,9 +41,10 @@
// go run ./cmd/seed-a1 -mode backfill-categories -mysql-dump path/to/descrybe_new.sql
// go run ./cmd/seed-a1 -mode backfill-attributes
//
// DATABASE_URL / -postgres required.
// DATABASE_URL / -postgres required (also loaded from monorepo-root .env via config.LoadDotEnv).
// Day-to-day: `npm run seed:a1` (reimport + category prompts + mapped category backfill). Maintainer snapshot: `npm run seed:a1:export`.
// recover-jobs is opt-in only and will reintroduce job history.
// Prefer admin UI Sync A1 on the company row (uses API DATABASE_URL; no CLI env needed).
package main
import (
@@ -61,6 +62,7 @@ import (
"compress/gzip"
"github.com/descrybe/descrybe-v2/apps/api/internal/billing"
"github.com/descrybe/descrybe-v2/apps/api/internal/config"
"github.com/google/uuid"
"github.com/jackc/pgx/v5"
"github.com/jackc/pgx/v5/pgconn"
@@ -82,6 +84,8 @@ type tableSpec struct {
}
func main() {
config.LoadDotEnv()
postgresURL := flag.String("postgres", os.Getenv("DATABASE_URL"), "Postgres URL (DATABASE_URL)")
mode := flag.String("mode", "reimport", "export | reimport | recover-jobs | apply-category-prompts | backfill-categories | backfill-attributes")
file := flag.String("file", "", "gzipped seed archive path (required for export/reimport)")
@@ -93,7 +97,7 @@ func main() {
flag.Parse()
if strings.TrimSpace(*postgresURL) == "" {
log.Fatal("-postgres / DATABASE_URL is required")
log.Fatal("-postgres / DATABASE_URL is required (export it, pass -postgres, or put DATABASE_URL in monorepo-root .env — same as the API; prefer admin Sync A1)")
}
companyID, err := uuid.Parse(strings.TrimSpace(*company))
if err != nil {
+4 -53
View File
@@ -12,6 +12,7 @@ import (
"time"
"github.com/descrybe/descrybe-v2/apps/api/internal/billing"
"github.com/descrybe/descrybe-v2/apps/api/internal/processing"
"github.com/google/uuid"
"github.com/jackc/pgx/v5/pgxpool"
)
@@ -419,65 +420,15 @@ func dumpSectionEnded(line, table string) bool {
}
func looksLikeTupleLine(line string) bool {
s := strings.TrimLeft(line, " \t")
return strings.HasPrefix(s, "(")
return processing.LooksLikeMySQLTupleLine(line)
}
func parseMySQLTupleFields(line string) []string {
return parseMySQLTupleFieldsN(line, 0)
return processing.ParseMySQLTupleFields(line)
}
// parseMySQLTupleFieldsN parses up to maxFields (0 = all) from the first (...) tuple on the line.
func parseMySQLTupleFieldsN(line string, maxFields int) []string {
start := strings.Index(line, "(")
if start < 0 {
return nil
}
body := line[start+1:]
var out []string
for i := 0; i < len(body); {
if maxFields > 0 && len(out) >= maxFields {
break
}
for i < len(body) && (body[i] == ' ' || body[i] == '\t' || body[i] == ',') {
i++
}
if i >= len(body) || body[i] == ')' {
break
}
if body[i] == '\'' {
i++
var b strings.Builder
for i < len(body) {
ch := body[i]
if ch == '\\' && i+1 < len(body) {
b.WriteByte(body[i+1])
i += 2
continue
}
if ch == '\'' {
if i+1 < len(body) && body[i+1] == '\'' {
b.WriteByte('\'')
i += 2
continue
}
i++
break
}
b.WriteByte(ch)
i++
}
out = append(out, b.String())
continue
}
j := i
for j < len(body) && body[j] != ',' && body[j] != ')' {
j++
}
out = append(out, strings.TrimSpace(body[i:j]))
i = j
}
return out
return processing.ParseMySQLTupleFieldsN(line, maxFields)
}
func parseDumpJobID(raw string) (uuid.UUID, error) {