Files
query-orchestration/internal/document/text/getPages_test.go
T

460 lines
9.9 KiB
Go
Raw Normal View History

package documenttext
import (
"context"
"encoding/json"
2025-04-23 17:51:44 +00:00
"fmt"
"io"
"os"
"strings"
"testing"
documenttypes "queryorchestration/internal/document/types"
textractmock "queryorchestration/mocks/textract"
"github.com/aws/aws-sdk-go-v2/service/textract"
"github.com/aws/aws-sdk-go-v2/service/textract/types"
"github.com/google/go-cmp/cmp"
"github.com/google/uuid"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/mock"
"github.com/stretchr/testify/require"
)
func TestGetBasePage(t *testing.T) {
ctx := context.Background()
cfg := DocTextConfig{}
svc := Service{
cfg: &cfg,
}
mockTextract := textractmock.NewMockTextractClient(t)
cfg.TextractClient = mockTextract
sent := 0
pdf, err := documenttypes.NewPDFFromReader(strings.NewReader(pdfHelloWorld))
require.NoError(t, err)
2025-04-23 17:51:44 +00:00
root, err := os.OpenRoot("../../../assets")
require.NoError(t, err)
textractBaseOut := getTextractFileResponse(t, "sampleGeneration/generated/helloWorld.gen", root)
mockTextract.EXPECT().
AnalyzeDocument(
mock.Anything,
mock.MatchedBy(func(in *textract.AnalyzeDocumentInput) bool {
if sent == 0 {
sent++
return true
}
return false
}),
mock.Anything,
).
Return(&textractBaseOut["base"][0], nil).Once()
text, err := svc.GetBasePage(ctx, pdf, 0)
require.NoError(t, err)
assert.EqualExportedValues(t, textractBaseOut["base"][0], text)
}
func TestGetDocumentText(t *testing.T) {
ctx := context.Background()
t.Run("Hello World", func(t *testing.T) {
2025-04-23 17:51:44 +00:00
t.Parallel()
cfg := DocTextConfig{}
svc := Service{
cfg: &cfg,
}
mockTextract := textractmock.NewMockTextractClient(t)
cfg.TextractClient = mockTextract
2025-04-23 17:51:44 +00:00
textractBaseOut, pdf, txtFile, close := getFiles(t, "sampleGeneration", "helloWorld")
defer close()
2025-04-23 17:51:44 +00:00
textExpectations(t, mockTextract, textractBaseOut)
text, err := svc.getDocumentText(ctx, pdf)
require.NoError(t, err)
2025-04-23 17:51:44 +00:00
assertReaders(t, txtFile, text)
})
t.Run("sampleOne - cnc_01-06", func(t *testing.T) {
2025-04-23 17:51:44 +00:00
t.Parallel()
cfg := DocTextConfig{}
svc := Service{
cfg: &cfg,
}
mockTextract := textractmock.NewMockTextractClient(t)
cfg.TextractClient = mockTextract
2025-04-23 17:51:44 +00:00
textractBaseOut, pdf, txtFile, close := getFiles(t, "sampleOne", "cnc_01-06")
defer close()
2025-04-23 17:51:44 +00:00
textExpectations(t, mockTextract, textractBaseOut)
2025-04-23 17:51:44 +00:00
text, err := svc.getDocumentText(ctx, pdf)
require.NoError(t, err)
2025-04-23 17:51:44 +00:00
assertReaders(t, txtFile, text)
})
t.Run("sampleOne - hn_0109", func(t *testing.T) {
t.Parallel()
cfg := DocTextConfig{}
svc := Service{
cfg: &cfg,
}
mockTextract := textractmock.NewMockTextractClient(t)
cfg.TextractClient = mockTextract
2025-04-23 17:51:44 +00:00
textractBaseOut, pdf, txtFile, close := getFiles(t, "sampleOne", "hn_0109")
defer close()
2025-04-23 17:51:44 +00:00
textExpectations(t, mockTextract, textractBaseOut)
text, err := svc.getDocumentText(ctx, pdf)
require.NoError(t, err)
assertReaders(t, txtFile, text)
})
2025-04-23 17:51:44 +00:00
t.Run("sampleOne - hn23-70", func(t *testing.T) {
t.Parallel()
cfg := DocTextConfig{}
svc := Service{
cfg: &cfg,
}
mockTextract := textractmock.NewMockTextractClient(t)
cfg.TextractClient = mockTextract
2025-04-23 17:51:44 +00:00
textractBaseOut, pdf, txtFile, close := getFiles(t, "sampleOne", "hn_23-70")
defer close()
2025-04-23 17:51:44 +00:00
textExpectations(t, mockTextract, textractBaseOut)
2025-04-23 17:51:44 +00:00
text, err := svc.getDocumentText(ctx, pdf)
require.NoError(t, err)
2025-04-23 17:51:44 +00:00
assertReaders(t, txtFile, text)
})
// t.Run("sampleOne - chc_1", func(t *testing.T) {
// t.Parallel()
// cfg := DocTextConfig{}
// svc := Service{
// cfg: &cfg,
// }
// mockTextract := textractmock.NewMockTextractClient(t)
// cfg.TextractClient = mockTextract
// textractBaseOut, pdf, txtFile, close := getFiles(t, "sampleOne", "chc_1")
// defer close()
// textExpectations(t, mockTextract, textractBaseOut)
// text, err := svc.getDocumentText(ctx, pdf)
// require.NoError(t, err)
// assertReaders(t, txtFile, text)
// })
2025-04-23 17:51:44 +00:00
}
2025-04-23 17:51:44 +00:00
func getFiles(t testing.TB, sample string, filename string) (map[string][]textract.AnalyzeDocumentOutput, *documenttypes.PDF, *os.File, func()) {
root, err := os.OpenRoot("../../../assets")
require.NoError(t, err)
root, err = root.OpenRoot(sample)
require.NoError(t, err)
2025-04-23 17:51:44 +00:00
textractFile, err := root.Open(fmt.Sprintf("generated/%s.gen", filename))
require.NoError(t, err)
defer textractFile.Close()
2025-04-23 17:51:44 +00:00
decoder := json.NewDecoder(textractFile)
var out map[string][]textract.AnalyzeDocumentOutput
err = decoder.Decode(&out)
require.NoError(t, err)
2025-04-23 17:51:44 +00:00
pdfFile, err := root.Open(fmt.Sprintf("%s.pdf", filename))
require.NoError(t, err)
2025-04-23 17:51:44 +00:00
pdf, err := documenttypes.NewPDFFromReader(pdfFile)
require.NoError(t, err)
txtFile, err := root.Open(fmt.Sprintf("%s.txt", filename))
require.NoError(t, err)
return out, pdf, txtFile, func() {
pdfFile.Close()
txtFile.Close()
}
}
2025-04-23 17:51:44 +00:00
func textExpectations(t testing.TB, mockTextract *textractmock.MockTextractClient, textractBaseOut map[string][]textract.AnalyzeDocumentOutput) {
sent := 0
tableIndex := 0
for i := range len(textractBaseOut["base"]) {
curr := i + tableIndex
mockTextract.EXPECT().
AnalyzeDocument(
mock.Anything,
mock.MatchedBy(func(in *textract.AnalyzeDocumentInput) bool {
2025-04-23 17:51:44 +00:00
if sent == curr {
sent++
return true
}
return false
}),
mock.Anything,
).
2025-04-23 17:51:44 +00:00
Return(&textractBaseOut["base"][i], nil).Once()
2025-04-23 17:51:44 +00:00
hasTable := false
for _, t := range textractBaseOut["base"][i].Blocks {
if t.BlockType == types.BlockTypeLayoutTable {
hasTable = true
break
}
}
2025-04-23 17:51:44 +00:00
if hasTable {
curr := i + tableIndex + 1
mockTextract.EXPECT().
AnalyzeDocument(
mock.Anything,
mock.MatchedBy(func(in *textract.AnalyzeDocumentInput) bool {
if sent == curr && in.FeatureTypes[0] == types.FeatureTypeTables {
sent++
return true
}
return false
}),
mock.Anything,
).
Return(&textractBaseOut["table"][tableIndex], nil).Once()
tableIndex++
}
}
}
func TestGetBlockMap(t *testing.T) {
svc := Service{}
t.Run("0 blocks", func(t *testing.T) {
blocks := []types.Block{}
outBlocks := svc.getBlockMap(blocks)
assert.Equal(t, map[uuid.UUID]types.Block{}, outBlocks)
})
t.Run("nil id blocks", func(t *testing.T) {
blocks := []types.Block{
{},
}
outBlocks := svc.getBlockMap(blocks)
assert.Equal(t, map[uuid.UUID]types.Block{}, outBlocks)
})
t.Run("1 block", func(t *testing.T) {
blockId := uuid.New()
blockIdStr := blockId.String()
blocks := []types.Block{
{
Id: &blockIdStr,
},
}
outBlocks := svc.getBlockMap(blocks)
assert.Equal(t, map[uuid.UUID]types.Block{
blockId: {
Id: &blockIdStr,
},
}, outBlocks)
})
t.Run("invalid block id", func(t *testing.T) {
blockIdStr := "blockId.String()"
blocks := []types.Block{
{
Id: &blockIdStr,
},
}
outBlocks := svc.getBlockMap(blocks)
assert.Equal(t, map[uuid.UUID]types.Block{}, outBlocks)
})
t.Run("2 blocks", func(t *testing.T) {
blockOneId := uuid.New()
blockOneIdStr := blockOneId.String()
blockTwoId := uuid.New()
blockTwoIdStr := blockTwoId.String()
blocks := []types.Block{
{
Id: &blockOneIdStr,
},
{
Id: &blockTwoIdStr,
},
}
outBlocks := svc.getBlockMap(blocks)
assert.Equal(t, map[uuid.UUID]types.Block{
blockOneId: {
Id: &blockOneIdStr,
},
blockTwoId: {
Id: &blockTwoIdStr,
},
}, outBlocks)
})
}
func TestMergeDocument(t *testing.T) {
svc := Service{}
t.Run("0 pages", func(t *testing.T) {
pages := []string{}
out, err := svc.mergeDocument(pages)
require.NoError(t, err)
outStr, err := io.ReadAll(out)
require.NoError(t, err)
assert.Equal(t, ``, string(outStr))
})
t.Run("1 page", func(t *testing.T) {
pages := []string{
"page one",
}
out, err := svc.mergeDocument(pages)
require.NoError(t, err)
outStr, err := io.ReadAll(out)
require.NoError(t, err)
assert.Equal(t, `Start of Page No. = 1
page one
`, string(outStr))
})
t.Run("2 pages", func(t *testing.T) {
pages := []string{
"page one",
"page two",
}
out, err := svc.mergeDocument(pages)
require.NoError(t, err)
outStr, err := io.ReadAll(out)
require.NoError(t, err)
assert.Equal(t, `Start of Page No. = 1
page one
Start of Page No. = 2
page two
`, string(outStr))
})
t.Run("3 pages", func(t *testing.T) {
pages := []string{
"page one",
"page two",
"page three",
}
out, err := svc.mergeDocument(pages)
require.NoError(t, err)
outStr, err := io.ReadAll(out)
require.NoError(t, err)
assert.Equal(t, `Start of Page No. = 1
page one
Start of Page No. = 2
page two
Start of Page No. = 3
page three
`, string(outStr))
})
}
2025-04-23 17:51:44 +00:00
func getTextractFileResponse(t testing.TB, filename string, root *os.Root) map[string][]textract.AnalyzeDocumentOutput {
file, err := root.Open(filename)
require.NoError(t, err)
defer file.Close()
decoder := json.NewDecoder(file)
var out map[string][]textract.AnalyzeDocumentOutput
err = decoder.Decode(&out)
require.NoError(t, err)
return out
}
2025-04-23 17:51:44 +00:00
func assertReaders(t testing.TB, expected io.Reader, actual io.Reader) {
actualBytes, err := io.ReadAll(actual)
require.NoError(t, err)
actualStr := string(actualBytes)
expectedBytes, err := io.ReadAll(expected)
require.NoError(t, err)
expectedStr := string(expectedBytes)
if !assert.Equal(t, expectedStr, actualStr) {
diff := cmp.Diff(expectedStr, actualStr)
t.Logf("Detailed diff (-expected +actual):\n%s", diff)
}
}
const pdfHelloWorld = `%PDF-1.4
%
1 0 obj
<<
/Type /Catalog
/Pages 2 0 R
/Version /1.4
>>
endobj
2 0 obj
<<
/Type /Pages
/Kids [3 0 R]
/Count 1
>>
endobj
3 0 obj
<<
/Type /Page
/Parent 2 0 R
/MediaBox [0 0 612 792]
/Resources <<
/Font <<
/F1 <<
/Type /Font
/Subtype /Type1
/BaseFont /Helvetica
>>
>>
>>
/Contents 4 0 R
>>
endobj
4 0 obj
<<
/Length
44
>>
stream
BT
/F1 24 Tf
100 700 Td
(Hello World!) Tj
ET
endstream
endobj
xref
0 5
0000000000 65535 f
0000000015 00000 n
0000000086 00000 n
0000000151 00000 n
0000000376 00000 n
trailer
<<
/Size 5
/Root 1 0 R
>>
startxref
472
%%EOF`