Github user dsmiley commented on a diff in the pull request:
https://github.com/apache/lucene-solr/pull/395#discussion_r194445760
--- Diff: solr/core/src/java/org/apache/solr/handler/loader/JsonLoader.java
---
@@ -249,14 +251,27 @@ public void handle(Map<String, Object> record, String
path) {
private SolrInputDocument buildDoc(Map<String, Object> m) {
SolrInputDocument result = new SolrInputDocument();
for (Map.Entry<String, Object> e : m.entrySet()) {
- if (e.getKey() == null) {// special case. JsonRecordReader emits
child docs with null key
+ if (entryIsChildDoc(e.getValue())) {// special case.
JsonRecordReader emits child docs with null key
if (e.getValue() instanceof List) {
List value = (List) e.getValue();
for (Object o : value) {
- if (o instanceof Map) result.addChildDocument(buildDoc((Map)
o));
+ if (o instanceof Map) {
+ if (anonChildDocFlag) {
+ result.addChildDocument(buildDoc((Map) o));
+ } else {
+ if(!result.containsKey(e.getKey())) {
+ result.setField(e.getKey(), new ArrayList<>(1));
--- End diff --
ok so you are trying to retain the semantic information that the input was
an array (potentially multi-valued) even if the particular input doc only had
one value. Maybe say more or that, hinting at why we even care. (why do we
care)? For regular field values, we don't at this level -- the schema holds
multiValue info so it's dealth with later. Again I'm liking my suggestion of
putting a virtual child doc field in the schema as it addresses the desire to
know this semantic info, plus I think it may add some consistency (avoids
special case you're doing here)
---
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
