[Pig-dev] [jira] Created: (PIG-526) Order of <key, value> pairs not preserved in MAP type.

[ https://issues.apache.org/jira/browse/PIG-526?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12647370#action_12647370 ]

Hong Tang commented on PIG-526:
-------------------------------

I suggest using LinkedHashMap instead.

[ https://issues.apache.org/jira/browse/PIG-526?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12647375#action_12647375 ]

Alan Gates commented on PIG-526:
--------------------------------

I don't know that pig promises any particular order for keys in a map. Do you believe we should be maintaining that order in memory, or is your concern that if we load and store the map it may be stored in a different order?

[ https://issues.apache.org/jira/browse/PIG-526?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12647385#action_12647385 ]

Hong Tang commented on PIG-526:
-------------------------------

Probably it is better for me to rephrase the issue as follows:

Need <key, value> order control on PIG MAP type.

For the lest, I would like to have the capability of keeping them sorted by key. Otherwise, it is more flexible to have PIG honor the ordering from the user-created Map object. The conceptual gap here is that: I can set a MAP datum using a SortedMap object when providing a tuple object; but after I write the tuple out and later read it back (through Writable interface), the iteration order of <key, value> pairs is completely changed.

[ https://issues.apache.org/jira/browse/PIG-526?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12647424#action_12647424 ]

Olga Natkovich commented on PIG-526:
------------------------------------

My concern with forcing this is performance consequences of maintaing order. I don't think Pig promises to the users any particular order

[ https://issues.apache.org/jira/browse/PIG-526?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12647432#action_12647432 ]

Hong Tang commented on PIG-526:
-------------------------------

I understand your concern. But just as I rephrased, the issue here is that PIG allows no control from user side to choose which concrete Map class to use when deserilizing the tuples.

Probably a good compromise is to allow user to specify which Map class should be used when performing Tuple deserialization.

[ https://issues.apache.org/jira/browse/PIG-526?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12647474#action_12647474 ]

Olga Natkovich commented on PIG-526:
------------------------------------

As of now Pig does not guarantee any order on the keys and we have not seen users complaining about it. So I am not sure that we need to do anything. Also, at the moment (until we move to 3.0 hopefully not soon.) we can't e breaking any existing interfaces.

[ https://issues.apache.org/jira/browse/PIG-526?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Hong Tang resolved PIG-526.
---------------------------

Resolution: Won't Fix