modules/text2vec-contextionary/vectorizer/schema_config_test.go

//                           _       _
// __      _____  __ ___   ___  __ _| |_ ___
// \ \ /\ / / _ \/ _` \ \ / / |/ _` | __/ _ \
//  \ V  V /  __/ (_| |\ V /| | (_| | ||  __/
//   \_/\_/ \___|\__,_| \_/ |_|\__,_|\__\___|
//
//  Copyright © 2016 - 2024 Weaviate B.V. All rights reserved.
//
//  CONTACT: [email protected]
//

package vectorizer

import (
	"context"
	"testing"

	"github.com/sirupsen/logrus"
	ltest "github.com/sirupsen/logrus/hooks/test"
	"github.com/stretchr/testify/assert"
	"github.com/stretchr/testify/require"
	"github.com/weaviate/weaviate/entities/models"
	"github.com/weaviate/weaviate/entities/schema"
)

func TestConfigValidator(t *testing.T) {
	t.Run("validate class names", func(t *testing.T) {
		type testCase struct {
			input     string
			valid     bool
			name      string
			vectorize bool
		}

		// for all test cases keep in mind that the word "carrot" is not present in
		// the fake c11y, but every other word is.
		//
		// Additionally, the word "the" is a stopword
		//
		// all inputs represent class names (!)
		tests := []testCase{
			// valid names
			{
				name:      "Single uppercase word present in the c11y",
				input:     "Car",
				valid:     true,
				vectorize: true,
			},
			{
				name:      "Single lowercase word present in the c11y, stored as uppercase",
				input:     "car",
				valid:     true,
				vectorize: true,
			},
			{
				name:      "combination of valid words starting with uppercase letter",
				input:     "CarGarage",
				valid:     true,
				vectorize: true,
			},
			{
				name:      "combination of valid words starting with lowercase letter, stored as uppercase",
				input:     "carGarage",
				valid:     true,
				vectorize: true,
			},
			{
				name:      "combination of valid words and stopwords, starting with uppercase",
				input:     "TheCarGarage",
				valid:     true,
				vectorize: true,
			},
			{
				name:      "combination of valid words and stopwords starting with lowercase letter, stored as uppercase",
				input:     "carTheGarage",
				valid:     true,
				vectorize: true,
			},

			// invalid names
			{
				name:      "Single uppercase word NOT present in the c11y",
				input:     "Carrot",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "Single lowercase word NOT present in the c11y",
				input:     "carrot",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "Single uppercase stopword",
				input:     "The",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "Single lowercase stopword",
				input:     "the",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of valid and invalid words, valid word first lowercased",
				input:     "potatoCarrot",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of valid and invalid words, valid word first uppercased",
				input:     "PotatoCarrot",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of valid and invalid words, invalid word first lowercased",
				input:     "carrotPotato",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of valid and invalid words, invalid word first uppercased",
				input:     "CarrotPotato",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of only stopwords, starting with lowercase",
				input:     "theThe",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of only stopwords, starting with uppercase",
				input:     "TheThe",
				valid:     false,
				vectorize: true,
			},

			// vectorize turned off
			{
				name:      "non-vectorized: combination of only stopwords, starting with uppercase",
				input:     "TheThe",
				valid:     true,
				vectorize: false,
			},
			{
				name:      "non-vectorized: excluded word",
				input:     "carrot",
				valid:     true,
				vectorize: false,
			},
		}

		for _, test := range tests {
			t.Run(test.name+" object class", func(t *testing.T) {
				class := &models.Class{
					Class: test.input,
					Properties: []*models.Property{{
						Name:         "dummyPropSoWeDontRunIntoAllNoindexedError",
						DataType:     schema.DataTypeText.PropString(),
						Tokenization: models.PropertyTokenizationWhitespace,
					}},
				}

				logger, _ := ltest.NewNullLogger()
				v := NewConfigValidator(&fakeRemote{}, logger)
				err := v.Do(context.Background(), class, nil, &fakeIndexChecker{
					vectorizeClassName: test.vectorize,
					propertyIndexed:    true,
				})
				assert.Equal(t, test.valid, err == nil)

				// only proceed if input was supposed to be valid
				if test.valid == false {
					return
				}
			})
		}
	})

	t.Run("validate property names", func(t *testing.T) {
		type testCase struct {
			input     string
			valid     bool
			name      string
			vectorize bool
		}

		// for all test cases keep in mind that the word "carrot" is not present in
		// the fake c11y, but every other word is
		//
		// all inputs represent property names (!)
		tests := []testCase{
			// valid names
			{
				name:      "Single uppercase word present in the c11y, stored as lowercase",
				input:     "Brand",
				valid:     true,
				vectorize: true,
			},
			{
				name:      "Single lowercase word present in the c11y",
				input:     "brand",
				valid:     true,
				vectorize: true,
			},
			{
				name:      "combination of valid words starting with uppercase letter, stored as lowercase",
				input:     "BrandGarage",
				valid:     true,
				vectorize: true,
			},
			{
				name:      "combination of valid words starting with lowercase letter",
				input:     "brandGarage",
				valid:     true,
				vectorize: true,
			},
			{
				name:      "combination of valid words and stop words starting with uppercase letter, stored as lowercase",
				input:     "TheGarage",
				valid:     true,
				vectorize: true,
			},
			{
				name:      "combination of valid words and stop words starting with lowercase letter",
				input:     "theGarage",
				valid:     true,
				vectorize: true,
			},

			// invalid names
			{
				name:      "Single uppercase word NOT present in the c11y",
				input:     "Carrot",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "Single lowercase word NOT present in the c11y",
				input:     "carrot",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "Single lowercase stop word",
				input:     "the",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of valid and invalid words, valid word first lowercased",
				input:     "potatoCarrot",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of valid and invalid words, valid word first uppercased",
				input:     "PotatoCarrot",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of valid and invalid words, invalid word first lowercased",
				input:     "carrotPotato",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of valid and invalid words, invalid word first uppercased",
				input:     "CarrotPotato",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of only stop words,  first lowercased",
				input:     "theThe",
				valid:     false,
				vectorize: true,
			},
			{
				name:      "combination of only stop words, first uppercased",
				input:     "TheThe",
				valid:     false,
				vectorize: true,
			},

			// without vectorizing
			{
				name:      "non-vectorizing: combination of only stop words, first uppercased",
				input:     "TheThe",
				valid:     true,
				vectorize: false,
			},
			{
				name:      "non-vectorizing: combination of only stop words, first uppercased",
				input:     "carrot",
				valid:     true,
				vectorize: false,
			},
		}

		for _, test := range tests {
			t.Run(test.name+" object class", func(t *testing.T) {
				class := &models.Class{
					Class: "ValidName",
					Properties: []*models.Property{{
						DataType:     schema.DataTypeText.PropString(),
						Tokenization: models.PropertyTokenizationWhitespace,
						Name:         test.input,
					}},
				}

				logger, _ := ltest.NewNullLogger()
				v := NewConfigValidator(&fakeRemote{}, logger)
				err := v.Do(context.Background(), class, nil, &fakeIndexChecker{
					vectorizePropertyName: test.vectorize,
					propertyIndexed:       true,
				})
				assert.Equal(t, test.valid, err == nil)
			})
		}
	})

	t.Run("all usable props no-indexed", func(t *testing.T) {
		t.Run("all schema vectorization turned off", func(t *testing.T) {
			class := &models.Class{
				Vectorizer: "text2vec-contextionary",
				Class:      "ValidName",
				Properties: []*models.Property{
					{
						DataType: []string{"text"},
						Name:     "description",
					},
					{
						DataType:     schema.DataTypeText.PropString(),
						Tokenization: models.PropertyTokenizationWhitespace,
						Name:         "name",
					},
					{
						DataType: []string{"int"},
						Name:     "amount",
					},
				},
			}

			logger, _ := ltest.NewNullLogger()
			v := NewConfigValidator(&fakeRemote{}, logger)
			err := v.Do(context.Background(), class, nil, &fakeIndexChecker{
				vectorizePropertyName: false,
				vectorizeClassName:    false,
				propertyIndexed:       false,
			})
			assert.NotNil(t, err)
		})
	})

	t.Run("with only array types", func(t *testing.T) {
		class := &models.Class{
			Vectorizer: "text2vec-contextionary",
			Class:      "ValidName",
			Properties: []*models.Property{
				{
					DataType: []string{"text[]"},
					Name:     "descriptions",
				},
				{
					DataType:     schema.DataTypeTextArray.PropString(),
					Tokenization: models.PropertyTokenizationWhitespace,
					Name:         "names",
				},
			},
		}

		logger, _ := ltest.NewNullLogger()
		v := NewConfigValidator(&fakeRemote{}, logger)
		err := v.Do(context.Background(), class, nil, &fakeIndexChecker{
			vectorizePropertyName: false,
			vectorizeClassName:    false,
			propertyIndexed:       true,
		})
		assert.Nil(t, err)
	})
}

func TestConfigValidator_RiskOfDuplicateVectors(t *testing.T) {
	type test struct {
		name          string
		in            *models.Class
		expectWarning bool
		indexChecker  *fakeIndexChecker
	}

	tests := []test{
		{
			name: "usable properties",
			in: &models.Class{
				Class: "ValidName",
				Properties: []*models.Property{
					{
						DataType: []string{string(schema.DataTypeText)},
						Name:     "textProp",
					},
				},
			},
			expectWarning: false,
			indexChecker: &fakeIndexChecker{
				vectorizePropertyName: false,
				vectorizeClassName:    true,
				propertyIndexed:       true,
			},
		},
		{
			name: "no properties",
			in: &models.Class{
				Class: "ValidName",
			},
			expectWarning: true,
			indexChecker: &fakeIndexChecker{
				vectorizePropertyName: false,
				vectorizeClassName:    true,
				propertyIndexed:       false,
			},
		},
		{
			name: "usable properties, but they are no-indexed",
			in: &models.Class{
				Class: "ValidName",
				Properties: []*models.Property{
					{
						DataType: []string{string(schema.DataTypeText)},
						Name:     "textProp",
					},
				},
			},
			expectWarning: true,
			indexChecker: &fakeIndexChecker{
				vectorizePropertyName: false,
				vectorizeClassName:    true,
				propertyIndexed:       false,
			},
		},
		{
			name: "only unusable properties",
			in: &models.Class{
				Class: "ValidName",
				Properties: []*models.Property{
					{
						DataType: []string{string(schema.DataTypeInt)},
						Name:     "intProp",
					},
				},
			},
			expectWarning: true,
			indexChecker: &fakeIndexChecker{
				vectorizePropertyName: false,
				vectorizeClassName:    true,
				propertyIndexed:       false,
			},
		},
	}

	for _, test := range tests {
		t.Run(test.name, func(t *testing.T) {
			logger, hook := ltest.NewNullLogger()
			v := NewConfigValidator(&fakeRemote{}, logger)
			err := v.Do(context.Background(), test.in, nil, test.indexChecker)
			require.Nil(t, err)

			entry := hook.LastEntry()
			if test.expectWarning {
				require.NotNil(t, entry)
				assert.Equal(t, logrus.WarnLevel, entry.Level)
			} else {
				assert.Nil(t, entry)
			}
		})
	}
}

type fakeIndexChecker struct {
	vectorizeClassName    bool
	vectorizePropertyName bool
	propertyIndexed       bool
}

func (f *fakeIndexChecker) VectorizeClassName() bool {
	return f.vectorizeClassName
}

func (f *fakeIndexChecker) VectorizePropertyName(propName string) bool {
	return f.vectorizePropertyName
}

func (f *fakeIndexChecker) PropertyIndexed(propName string) bool {
	return f.propertyIndexed
}

// Every word in this fake c11y remote client is present except for the word
// Carrot which is not present
type fakeRemote struct{}

func (f *fakeRemote) IsWordPresent(ctx context.Context, word string) (bool, error) {
	if word == "carrot" || word == "the" {
		return false, nil
	}
	return true, nil
}

func (f *fakeRemote) IsStopWord(ctx context.Context, word string) (bool, error) {
	return word == "the", nil
}