Hisoka-X commented on code in PR #9673:
URL: https://github.com/apache/seatunnel/pull/9673#discussion_r2284318174
##########
seatunnel-transforms-v2/src/main/java/org/apache/seatunnel/transform/nlpmodel/embedding/EmbeddingTransform.java:
##########
@@ -180,33 +187,49 @@ public void open() {
}
}
- private void initOutputFields(SeaTunnelRowType inputRowType, Map<String,
String> fields) {
+ private void initOutputFields(SeaTunnelRowType inputRowType,
ReadonlyConfig config) {
+ Map<Integer, FieldSpec> fieldSpecMap = new HashMap<>();
List<String> fieldNames = new ArrayList<>();
- List<Integer> fieldOriginalIndexes = new ArrayList<>();
- for (Map.Entry<String, String> field : fields.entrySet()) {
- String srcField = field.getValue();
+ Map<String, Object> fieldsConfig =
+ config.get(EmbeddingTransformConfig.VECTORIZATION_FIELDS);
+ if (fieldsConfig == null || fieldsConfig.isEmpty()) {
+ throw new IllegalArgumentException("vectorization_fields
configuration is required");
+ }
+
+ for (Map.Entry<String, Object> field : fieldsConfig.entrySet()) {
+ FieldSpec fieldSpec = new FieldSpec(field);
+ String srcField = fieldSpec.getFieldName();
int srcFieldIndex;
try {
srcFieldIndex = inputRowType.indexOf(srcField);
} catch (IllegalArgumentException e) {
throw
TransformCommonError.cannotFindInputFieldError(getPluginName(), srcField);
}
+ if (fieldSpec.isMultimodalField()) {
+ isMultimodalFields = true;
+ }
+ fieldSpecMap.put(srcFieldIndex, fieldSpec);
fieldNames.add(field.getKey());
- fieldOriginalIndexes.add(srcFieldIndex);
}
+ this.fieldSpecMap = fieldSpecMap;
this.fieldNames = fieldNames;
- this.fieldOriginalIndexes = fieldOriginalIndexes;
}
@Override
protected Object[] getOutputFieldValues(SeaTunnelRowAccessor inputRow) {
tryOpen();
try {
- Object[] fieldArray = new Object[fieldOriginalIndexes.size()];
- for (int i = 0; i < fieldOriginalIndexes.size(); i++) {
- fieldArray[i] = inputRow.getField(fieldOriginalIndexes.get(i));
+ Set<Integer> fieldOriginalIndexes = fieldSpecMap.keySet();
+ Object[] fieldValues = new Object[fieldOriginalIndexes.size()];
+ List<ByteBuffer> vectorization;
+ int i = 0;
+ for (Integer fieldOriginalIndex : fieldOriginalIndexes) {
+ FieldSpec fieldSpec = fieldSpecMap.get(fieldOriginalIndex);
+ Object value = inputRow.getField(fieldOriginalIndex);
+ fieldValues[i++] =
+ isMultimodalFields ? new
MultimodalFieldValue(fieldSpec, value) : value;
Review Comment:
In the binary format scenario, the data contained in each row may be part of
the complete file. We need to cache it in the transform and perform embedding
after obtaining the complete file. Please refer to
https://github.com/apache/seatunnel/blob/dev/seatunnel-connectors-v2/connector-file/connector-file-base/src/main/java/org/apache/seatunnel/connectors/seatunnel/file/source/reader/BinaryReadStrategy.java#L132
and
https://github.com/apache/seatunnel/blob/dev/seatunnel-connectors-v2/connector-file/connector-file-base/src/main/java/org/apache/seatunnel/connectors/seatunnel/file/sink/writer/BinaryWriteStrategy.java#L78
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
