mythrocks commented on a change in pull request #807:
URL: https://github.com/apache/hive/pull/807#discussion_r434898293
##########
File path: serde/src/java/org/apache/hadoop/hive/serde2/avro/InstanceCache.java
##########
@@ -51,20 +51,19 @@ public Instance retrieve(SeedObject hv) throws
AvroSerdeException {
* Retrieve (or create if it doesn't exist) the correct Instance for this
* SeedObject using 'seenSchemas' to resolve circular references
*/
- public synchronized Instance retrieve(SeedObject hv,
- Set<SeedObject> seenSchemas) throws AvroSerdeException {
+ public Instance retrieve(SeedObject hv, Set<SeedObject> seenSchemas)
+ throws AvroSerdeException {
if(LOG.isDebugEnabled()) LOG.debug("Checking for hv: " + hv.toString());
if(cache.containsKey(hv)) {
Review comment:
Ok, I think I have it.
[`InstanceCache::retrieve()`](https://github.com/apache/hive/blob/f37c5de6c32b9395d1b34fa3c02ed06d1bfbf6eb/serde/src/java/org/apache/hadoop/hive/serde2/avro/InstanceCache.java#L66)
modifies the cache here:
```java
Instance instance = makeInstance(hv, seenSchemas);
cache.put(hv, instance);
return instance;
```
We would like to use `computeIfAbsent()` instead:
```java
cache.computeIfAbsent(hv, ()->makeInstance(hv, seenSchemas));
```
If `makeInstance()` ends up modifying `cache`, (e.g. by calling
`retrieve()`), that would cause undefined behaviour in`computeIfAbsent()`.
[`SchemaToTypeInfo`](https://github.com/apache/hive/blob/f37c5de6c32b9395d1b34fa3c02ed06d1bfbf6eb/serde/src/java/org/apache/hadoop/hive/serde2/avro/SchemaToTypeInfo.java#L116)
implements `InstanceCache::makeInstance()` as follows:
```java
static InstanceCache<Schema, TypeInfo> typeInfoCache = new
InstanceCache<Schema, TypeInfo>() {
@Override
protected TypeInfo makeInstance(Schema s,
Set<Schema> seenSchemas)
throws AvroSerdeException {
return generateTypeInfoWorker(s,
seenSchemas); // <---- HERE!
}
};
```
Please note that
[`SchemaToTypeInfo::generateTypeInfo()`](https://github.com/apache/hive/blob/f37c5de6c32b9395d1b34fa3c02ed06d1bfbf6eb/serde/src/java/org/apache/hadoop/hive/serde2/avro/SchemaToTypeInfo.java#L186)
happens to call `InstanceCache::retrieve()`:
```java
public static TypeInfo generateTypeInfo(Schema schema,
Set<Schema> seenSchemas) throws AvroSerdeException {
// ...
return typeInfoCache.retrieve(schema, seenSchemas);
}
```
Hence, the recursive call. :/
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
