//go:build aws //go:generate go run -tags=aws ./textdetection.go package main import ( "context" "encoding/json" "fmt" "log/slog" "os" documenttext "queryorchestration/internal/document/text" documenttypes "queryorchestration/internal/document/types" "queryorchestration/internal/serviceconfig" "queryorchestration/internal/serviceconfig/aws" "queryorchestration/internal/serviceconfig/objectstore" "queryorchestration/internal/serviceconfig/queue/querysync" "queryorchestration/internal/serviceconfig/textract" "sync" "time" awstextract "github.com/aws/aws-sdk-go-v2/service/textract" "github.com/aws/aws-sdk-go-v2/service/textract/types" ) type Config struct { serviceconfig.BaseConfig textract.TextractConfig querysync.QuerySyncConfig objectstore.ObjectStoreConfig } func main() { ctx := context.Background() if len(os.Args) != 3 { fmt.Println("Error: expected at least two arguments") fmt.Println("Usage:", os.Args[0], "", "") os.Exit(1) } sampleName := os.Args[1] name := os.Args[2] slog.Info("arguments", "sample_name", sampleName, "file_name", name) cfg := &Config{} profile := aws.Profile(os.Getenv("AWS_PROFILE")) if profile == "" { slog.Error("profile not available") os.Exit(1) } err := cfg.SetTextractClientWithProfile(ctx, profile) if err != nil { slog.Error("error creating client", "error", err) os.Exit(1) } root, err := os.OpenRoot("assets") if err != nil { slog.Error("error opening root", "error", err) os.Exit(1) } root, err = root.OpenRoot(sampleName) if err != nil { slog.Error("error opening sample", "error", err) os.Exit(1) } pdfFile, err := root.Open(fmt.Sprintf("%s.pdf", name)) if err != nil { slog.Error("error opening pdf file", "error", err) os.Exit(1) } pdf, err := documenttypes.NewPDFFromReader(pdfFile) if err != nil { slog.Error("error creating pdf file", "error", err) os.Exit(1) } textSvc := documenttext.New(cfg) count, err := pdf.GetPageCount(ctx) if err != nil { slog.Error("error getting page count", "error", err) os.Exit(1) } slog.Info("detecting text", "page_count", count) startTime := time.Now() allResults := map[string][]awstextract.AnalyzeDocumentOutput{} allResults["base"] = make([]awstextract.AnalyzeDocumentOutput, count) var wg sync.WaitGroup for i := range count { wg.Add(1) go func() { slog.Info("getting page", "index", i) result, err := textSvc.GetBasePage(ctx, pdf, i) if err != nil { slog.Error("error detecting text", "error", err) os.Exit(1) } allResults["base"][i] = result wg.Done() }() } wg.Wait() allResults["table"] = []awstextract.AnalyzeDocumentOutput{} for i, result := range allResults["base"] { hasTable := false for _, block := range result.Blocks { if block.BlockType == types.BlockTypeLayoutTable { hasTable = true } } if hasTable { slog.Info("getting table", "index", i) pageBytes, err := pdf.GetPageAsPNG(ctx, i) if err != nil { slog.Error("error getting page", "error", err) os.Exit(1) } res, err := cfg.GetTextractClient().AnalyzeDocument(ctx, &awstextract.AnalyzeDocumentInput{ Document: &types.Document{ Bytes: pageBytes, }, FeatureTypes: []types.FeatureType{ types.FeatureTypeTables, }, }) if err != nil { slog.Error("error detecting table", "error", err) os.Exit(1) } allResults["table"] = append(allResults["table"], *res) } } generatedBase := "generated" if _, err := root.Stat(generatedBase); os.IsNotExist(err) { err := root.Mkdir(generatedBase, 0755) if err != nil { slog.Error("error creating directory", "error", err) os.Exit(1) } slog.Info("directory created", "directory", generatedBase) } jsonContent, err := json.MarshalIndent(allResults, "", "\t") if err != nil { slog.Error("error marshaling json", "error", err) os.Exit(1) } outputFilename := fmt.Sprintf("%s/%s.gen", generatedBase, name) file, err := root.Create(outputFilename) if err != nil { slog.Error("error creating file", "error", err) os.Exit(1) } defer file.Close() _, err = file.Write(jsonContent) if err != nil { slog.Error("error writing file", "error", err) os.Exit(1) } slog.Info("extracted text", "output", outputFilename, "time", time.Since(startTime)) }