|
| 1 | +package tools.jackson.dataformat.avro.schemaev; |
| 2 | + |
| 3 | +import org.junit.jupiter.api.Test; |
| 4 | + |
| 5 | +import com.fasterxml.jackson.annotation.JsonInclude; |
| 6 | + |
| 7 | +import tools.jackson.core.exc.StreamReadException; |
| 8 | + |
| 9 | +import tools.jackson.dataformat.avro.*; |
| 10 | + |
| 11 | +import static org.junit.jupiter.api.Assertions.*; |
| 12 | + |
| 13 | +/** |
| 14 | + * Test for issue #275: Avro backward compatibility when adding fields with default values. |
| 15 | + * <p> |
| 16 | + * This test demonstrates that Jackson Avro DOES support backward compatibility correctly, |
| 17 | + * but users MUST use the {@code withReaderSchema()} method to enable schema resolution. |
| 18 | + * <p> |
| 19 | + * The key insight is that Avro binary format does not include schema metadata in the |
| 20 | + * serialized data. Therefore, when reading data that was written with schema A using |
| 21 | + * schema B, the library needs to be explicitly told about both schemas through the |
| 22 | + * {@code withReaderSchema()} API. |
| 23 | + * <p> |
| 24 | + * Common mistake: Trying to read old data with a new schema directly leads to |
| 25 | + * "Unexpected end-of-input" errors because the parser tries to read fields that |
| 26 | + * don't exist in the binary data. |
| 27 | + * <p> |
| 28 | + * Correct usage pattern: |
| 29 | + * <pre> |
| 30 | + * // Write with old schema |
| 31 | + * AvroSchema writerSchema = mapper.schemaFrom(OLD_SCHEMA_JSON); |
| 32 | + * byte[] data = mapper.writer(writerSchema).writeValueAsBytes(object); |
| 33 | + * |
| 34 | + * // Read with new schema (that has additional fields with defaults) |
| 35 | + * AvroSchema readerSchema = mapper.schemaFrom(NEW_SCHEMA_JSON); |
| 36 | + * AvroSchema resolved = writerSchema.withReaderSchema(readerSchema); |
| 37 | + * MyObject result = mapper.readerFor(MyObject.class) |
| 38 | + * .with(resolved) // Use resolved schema, not readerSchema directly! |
| 39 | + * .readValue(data); |
| 40 | + * </pre> |
| 41 | + */ |
| 42 | +public class Evolution275Test extends AvroTestBase |
| 43 | +{ |
| 44 | + // Original schema with 8 fields (simulating the issue scenario) |
| 45 | + static String SCHEMA_V1_JSON = aposToQuotes("{\n"+ |
| 46 | + " 'type':'record',\n"+ |
| 47 | + " 'name':'Employee',\n"+ |
| 48 | + " 'fields':[\n"+ |
| 49 | + " { 'name':'code', 'type':'string' },\n"+ |
| 50 | + " { 'name':'countryCode', 'type':'string' },\n"+ |
| 51 | + " { 'name':'createdBy', 'type':'string' },\n"+ |
| 52 | + " { 'name':'createdDate', 'type':'string' },\n"+ |
| 53 | + " { 'name':'id', 'type':'long' },\n"+ |
| 54 | + " { 'name':'lastModifiedBy', 'type':'string' },\n"+ |
| 55 | + " { 'name':'lastModifiedDate', 'type':'string' },\n"+ |
| 56 | + " { 'name':'name', 'type':'string' }\n"+ |
| 57 | + " ]\n"+ |
| 58 | + "}\n"); |
| 59 | + |
| 60 | + // Updated schema adding a 9th field with null default at the end |
| 61 | + static String SCHEMA_V2_JSON = aposToQuotes("{\n"+ |
| 62 | + " 'type':'record',\n"+ |
| 63 | + " 'name':'Employee',\n"+ |
| 64 | + " 'fields':[\n"+ |
| 65 | + " { 'name':'code', 'type':'string' },\n"+ |
| 66 | + " { 'name':'countryCode', 'type':'string' },\n"+ |
| 67 | + " { 'name':'createdBy', 'type':'string' },\n"+ |
| 68 | + " { 'name':'createdDate', 'type':'string' },\n"+ |
| 69 | + " { 'name':'id', 'type':'long' },\n"+ |
| 70 | + " { 'name':'lastModifiedBy', 'type':'string' },\n"+ |
| 71 | + " { 'name':'lastModifiedDate', 'type':'string' },\n"+ |
| 72 | + " { 'name':'name', 'type':'string' },\n"+ |
| 73 | + " { 'name':'phone', 'type':['null', 'string'], 'default':null }\n"+ |
| 74 | + " ]\n"+ |
| 75 | + "}\n"); |
| 76 | + |
| 77 | + // Simpler test with just 2 fields + new field with null default |
| 78 | + static String SCHEMA_SIMPLE_V1_JSON = aposToQuotes("{\n"+ |
| 79 | + " 'type':'record',\n"+ |
| 80 | + " 'name':'SimpleRecord',\n"+ |
| 81 | + " 'fields':[\n"+ |
| 82 | + " { 'name':'id', 'type':'int' },\n"+ |
| 83 | + " { 'name':'name', 'type':'string' }\n"+ |
| 84 | + " ]\n"+ |
| 85 | + "}\n"); |
| 86 | + |
| 87 | + static String SCHEMA_SIMPLE_V2_JSON = aposToQuotes("{\n"+ |
| 88 | + " 'type':'record',\n"+ |
| 89 | + " 'name':'SimpleRecord',\n"+ |
| 90 | + " 'fields':[\n"+ |
| 91 | + " { 'name':'id', 'type':'int' },\n"+ |
| 92 | + " { 'name':'name', 'type':'string' },\n"+ |
| 93 | + " { 'name':'phone', 'type':['null', 'string'], 'default':null }\n"+ |
| 94 | + " ]\n"+ |
| 95 | + "}\n"); |
| 96 | + |
| 97 | + @JsonInclude(JsonInclude.Include.NON_NULL) |
| 98 | + public static class Employee { |
| 99 | + public String code; |
| 100 | + public String countryCode; |
| 101 | + public String createdBy; |
| 102 | + public String createdDate; |
| 103 | + public long id; |
| 104 | + public String lastModifiedBy; |
| 105 | + public String lastModifiedDate; |
| 106 | + public String name; |
| 107 | + public String phone; |
| 108 | + |
| 109 | + protected Employee() { } |
| 110 | + |
| 111 | + public Employee(String code, String countryCode, String createdBy, |
| 112 | + String createdDate, long id, String lastModifiedBy, |
| 113 | + String lastModifiedDate, String name) { |
| 114 | + this.code = code; |
| 115 | + this.countryCode = countryCode; |
| 116 | + this.createdBy = createdBy; |
| 117 | + this.createdDate = createdDate; |
| 118 | + this.id = id; |
| 119 | + this.lastModifiedBy = lastModifiedBy; |
| 120 | + this.lastModifiedDate = lastModifiedDate; |
| 121 | + this.name = name; |
| 122 | + } |
| 123 | + } |
| 124 | + |
| 125 | + @JsonInclude(JsonInclude.Include.NON_NULL) |
| 126 | + public static class SimpleRecord { |
| 127 | + public int id; |
| 128 | + public String name; |
| 129 | + public String phone; |
| 130 | + |
| 131 | + protected SimpleRecord() { } |
| 132 | + |
| 133 | + public SimpleRecord(int id, String name) { |
| 134 | + this.id = id; |
| 135 | + this.name = name; |
| 136 | + } |
| 137 | + } |
| 138 | + |
| 139 | + private final AvroMapper MAPPER = newMapper(); |
| 140 | + |
| 141 | + @Test |
| 142 | + public void testSimpleAddNullableFieldWithDefault() throws Exception |
| 143 | + { |
| 144 | + final AvroSchema srcSchema = MAPPER.schemaFrom(SCHEMA_SIMPLE_V1_JSON); |
| 145 | + final AvroSchema dstSchema = MAPPER.schemaFrom(SCHEMA_SIMPLE_V2_JSON); |
| 146 | + final AvroSchema xlate = srcSchema.withReaderSchema(dstSchema); |
| 147 | + |
| 148 | + // Write data using old schema (without phone field) |
| 149 | + byte[] avro = MAPPER.writer(srcSchema).writeValueAsBytes(new SimpleRecord(1, "Alice")); |
| 150 | + |
| 151 | + // Read using new schema (with phone field defaulting to null) |
| 152 | + // This should NOT throw "Unexpected end-of-input in FIELD_NAME" |
| 153 | + SimpleRecord result = MAPPER.readerFor(SimpleRecord.class) |
| 154 | + .with(xlate) |
| 155 | + .readValue(avro); |
| 156 | + |
| 157 | + assertEquals(1, result.id); |
| 158 | + assertEquals("Alice", result.name); |
| 159 | + assertNull(result.phone); // Should use default value |
| 160 | + } |
| 161 | + |
| 162 | + // This test demonstrates INCORRECT usage: trying to read data serialized with an old schema |
| 163 | + // using a new schema directly, without calling withReaderSchema(). |
| 164 | + // This is expected to fail because Avro binary format doesn't include schema metadata, |
| 165 | + // so the reader can't know the data was written with a different schema. |
| 166 | + // Users MUST call withReaderSchema() when reading data written with a different schema. |
| 167 | + @Test |
| 168 | + public void testSimpleAddNullableFieldWithDefaultWrongUsage() throws Exception |
| 169 | + { |
| 170 | + final AvroSchema srcSchema = MAPPER.schemaFrom(SCHEMA_SIMPLE_V1_JSON); |
| 171 | + final AvroSchema dstSchema = MAPPER.schemaFrom(SCHEMA_SIMPLE_V2_JSON); |
| 172 | + |
| 173 | + // Write data using old schema (without phone field) |
| 174 | + byte[] avro = MAPPER.writer(srcSchema).writeValueAsBytes(new SimpleRecord(1, "Alice")); |
| 175 | + |
| 176 | + // INCORRECT: Try to read with new schema directly without using withReaderSchema |
| 177 | + // This triggers EOF error because the reader expects to find the phone field in binary data |
| 178 | + // but the data doesn't contain it. |
| 179 | + StreamReadException thrown = assertThrows(StreamReadException.class, () -> { |
| 180 | + MAPPER.readerFor(SimpleRecord.class) |
| 181 | + .with(dstSchema) // Using dstSchema directly instead of xlate |
| 182 | + .readValue(avro); |
| 183 | + }); |
| 184 | + |
| 185 | + verifyException(thrown, "Unexpected end-of-input in PROPERTY_NAME"); |
| 186 | + } |
| 187 | + |
| 188 | + @Test |
| 189 | + public void testAddNullableFieldWithDefault() throws Exception |
| 190 | + { |
| 191 | + final AvroSchema srcSchema = MAPPER.schemaFrom(SCHEMA_V1_JSON); |
| 192 | + final AvroSchema dstSchema = MAPPER.schemaFrom(SCHEMA_V2_JSON); |
| 193 | + final AvroSchema xlate = srcSchema.withReaderSchema(dstSchema); |
| 194 | + |
| 195 | + // Write data using old schema (without phone field) |
| 196 | + Employee emp = new Employee("EMP001", "US", "admin", "2024-01-01", |
| 197 | + 123L, "admin", "2024-01-01", "John Doe"); |
| 198 | + byte[] avro = MAPPER.writer(srcSchema).writeValueAsBytes(emp); |
| 199 | + |
| 200 | + // Read using new schema (with phone field defaulting to null) |
| 201 | + // This should NOT throw "Unexpected end-of-input in FIELD_NAME" |
| 202 | + Employee result = MAPPER.readerFor(Employee.class) |
| 203 | + .with(xlate) |
| 204 | + .readValue(avro); |
| 205 | + |
| 206 | + assertEquals("EMP001", result.code); |
| 207 | + assertEquals("US", result.countryCode); |
| 208 | + assertEquals("admin", result.createdBy); |
| 209 | + assertEquals("2024-01-01", result.createdDate); |
| 210 | + assertEquals(123L, result.id); |
| 211 | + assertEquals("admin", result.lastModifiedBy); |
| 212 | + assertEquals("2024-01-01", result.lastModifiedDate); |
| 213 | + assertEquals("John Doe", result.name); |
| 214 | + assertNull(result.phone); // Should use default value |
| 215 | + } |
| 216 | +} |
0 commit comments