query-orchestration/internal/document/text/pageText.go

package documenttext

import (
	"context"
	"fmt"
	"log/slog"
	"strings"

	documenttypes "queryorchestration/internal/document/types"

	"github.com/aws/aws-sdk-go-v2/service/textract"
	"github.com/aws/aws-sdk-go-v2/service/textract/types"
	"github.com/google/uuid"
)

func (s *Service) getPageText(ctx context.Context, document documenttypes.File, index int) (string, error) {
	res, err := s.GetBasePage(ctx, document, index)
	if err != nil {
		return "", err
	}

	var pageBlock *types.Block
	for _, block := range res.Blocks {
		if block.BlockType == types.BlockTypePage {
			pageBlock = &block
			break
		}
	}
	if pageBlock == nil {
		return "", fmt.Errorf("no page block found for index: %d", index)
	}

	blockMap := s.getBlockMap(res.Blocks)

	lines, specialFeatures, err := s.getPageElements(ctx, document, index, *pageBlock, blockMap)
	if err != nil {
		return "", err
	}

	return s.buildPageText(ctx, lines, blockMap, index, specialFeatures)
}

func (s *Service) getPageElements(ctx context.Context, document documenttypes.File, index int, pageBlock types.Block, blockMap map[uuid.UUID]types.Block) ([]uuid.UUID, specialisedFeatures, error) {
	directChildren := s.getDirectChildren(pageBlock, blockMap)

	hasTable := false
	for _, id := range directChildren {
		if blockMap[id].BlockType == types.BlockTypeLayoutTable {
			hasTable = true
		}
	}

	features := []types.FeatureType{}
	if hasTable {
		features = append(features, types.FeatureTypeTables)
		// features = append(features, types.FeatureTypeForms)
	}

	// Detecting Checkboxes -
	// if has tables
	// For a page with no tables, we look for 3 out of 4 of the following conditions
	//     has_exhibit→ 'exhibit' is on the page
	//     has_benefit→ one of ['benefit plan', 'benefit plan(s)', 'applicable benefit', 'benefits'] is on the page
	//     has_plans→ one of ['plan(s):', 'plan(s)/program(s):', 'plans:', 'plan/program:', 'plan(s)', 'plans'] is on the page
	//     has_services → one of ['services:', 'service:', 'service(s):', 'services', 'service(s)'] is on the page

	if len(features) == 0 {
		return directChildren, specialisedFeatures{}, nil
	}

	featuresOut, err := s.getSpecialisedFeatures(ctx, document, index, features)
	if err != nil {
		return nil, specialisedFeatures{}, err
	}

	return directChildren, featuresOut, nil
}

type buildParams struct {
	numOfSignatures int
	currentTable    int
}

func (s *Service) buildPageText(ctx context.Context, lines []uuid.UUID, blockMap map[uuid.UUID]types.Block, index int, specialFeatures specialisedFeatures) (string, error) {
	var builder strings.Builder
	params := buildParams{
		numOfSignatures: 0,
		currentTable:    0,
	}

	for _, id := range lines {
		err := s.addComponentText(ctx, blockMap[id], blockMap, specialFeatures, &params, &builder)
		if err != nil {
			return "", err
		}
	}

	for params.currentTable < len(specialFeatures.Tables) {
		id := specialFeatures.Tables[params.currentTable]
		err := s.addComponentText(ctx, specialFeatures.BlockMap[id], blockMap, specialFeatures, &params, &builder)
		if err != nil {
			return "", err
		}
	}

	if params.numOfSignatures > 0 {
		builder.WriteString(fmt.Sprintf("\nThis page has %d signature.\n", params.numOfSignatures))
	}

	return builder.String(), nil
}

func (s *Service) addComponentText(ctx context.Context, block types.Block, blockMap map[uuid.UUID]types.Block, specialFeatures specialisedFeatures, params *buildParams, builder *strings.Builder) error {
	switch block.BlockType {
	case types.BlockTypeLayoutText, types.BlockTypeLayoutTitle, types.BlockTypeLayoutHeader, types.BlockTypeLayoutFooter, types.BlockTypeLayoutSectionHeader, types.BlockTypeLayoutPageNumber, types.BlockTypeLayoutList, types.BlockTypeLayoutFigure, types.BlockTypeLayoutKeyValue:
		for _, id := range s.getChildIds(block) {
			err := s.addComponentText(ctx, blockMap[id], blockMap, specialFeatures, params, builder)
			if err != nil {
				return err
			}
		}
	case types.BlockTypeLine:
		_, err := builder.WriteString(fmt.Sprintf("%s\n", *block.Text))
		if err != nil {
			slog.Error("unable to write string", "error", err)
			return err
		}
	case types.BlockTypeWord:
		_, err := builder.WriteString(fmt.Sprintf(" %s", *block.Text))
		if err != nil {
			slog.Error("unable to write string", "error", err)
			return err
		}
	case types.BlockTypeSignature:
		params.numOfSignatures++
	case types.BlockTypeLayoutTable, types.BlockTypeTable:
		if block.BlockType == types.BlockTypeLayoutTable && len(specialFeatures.Tables) <= params.currentTable {
			for _, child := range s.getChildIds(block) {
				err := s.addComponentText(ctx, blockMap[child], blockMap, specialFeatures, params, builder)
				if err != nil {
					return err
				}
			}
			params.currentTable++
			break
		}
		tableId := specialFeatures.Tables[params.currentTable]
		params.currentTable++
		tableBlock := specialFeatures.BlockMap[tableId]
		err := s.addTable(ctx, tableBlock, specialFeatures.BlockMap, builder)
		if err != nil {
			return err
		}
	default:
		slog.Debug("uncaught component", "type", block.BlockType)
	}

	return nil
}

func (s *Service) GetBasePage(ctx context.Context, doc documenttypes.File, index int) (textract.AnalyzeDocumentOutput, error) {
	pageBytes, err := doc.GetPageAsPNG(ctx, index)
	if err != nil {
		return textract.AnalyzeDocumentOutput{}, err
	}

	res, err := s.cfg.GetTextractClient().AnalyzeDocument(ctx, &textract.AnalyzeDocumentInput{
		Document: &types.Document{
			Bytes: pageBytes,
		},
		FeatureTypes: []types.FeatureType{
			types.FeatureTypeLayout,
			types.FeatureTypeSignatures,
		},
	})
	if err != nil {
		return textract.AnalyzeDocumentOutput{}, err
	}

	return *res, err
}