[jira] [Commented] (HIVE-17674) grep TODO HIVE-15212.17.patch |wc - l = 49

["Sergey Shelukhin (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-17674?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16190226#comment-16190226 ] 

Sergey Shelukhin commented on HIVE-17674:
-----------------------------------------

Some need fixing after merge, some removal, some can be converted to notes.
The ones that document the shared limitation of the old path and MM path can be kept as is.
I'll look at it at.

> grep TODO HIVE-15212.17.patch |wc - l = 49
> ------------------------------------------
>
>                 Key: HIVE-17674
>                 URL: https://issues.apache.org/jira/browse/HIVE-17674
>             Project: Hive
>          Issue Type: Sub-task
>          Components: Transactions
>            Reporter: Eugene Koifman
>
> grep TODO HIVE-15212.17.patch 
> +  // TODO The following two utility methods can be moved to AcidUtils once no class in metastore is relying on them,
> +  // TODO: CTAS for MM may currently be broken. It used to work. See the old code and why isCtas isn't used?
> +    // TODO: Setting these separately is a very hairy issue in certain combinations, since we
> +    // TODO: this doesn't handle list bucketing properly. Does the original exim do that?
> +      // TODO: support this? we only bail because it's a PITA and hardly anyone seems to care.
> +    // TODO: support this?
> +    // TODO: support this?
> +          // TODO: why is it stored in both?
> +      // TODO: this doesn't work... the path for writer and reader mismatch
> -            // TODO: In a follow-up to HIVE-1361, we should refactor loadDynamicPartitions
> +    // TODO: In a follow-up to HIVE-1361, we should refactor loadDynamicPartitions
> +      // TODO: this particular bit will not work for MM tables, as there can be multiple
> +        // TODO: not clear why two if conditions are different. Preserve the existing logic for now.
> +      // TODO: not clear why two if conditions are different. Preserve the existing logic for now.
> +  // TODO: not clear why two if conditions are different. Preserve the existing logic for now.
> +  // TODO: not clear why two if conditions are different. Preserve the existing logic for now.
> +      // TODO: do we also need to do this when creating an empty partition from select?
> +    // TODO: see HIVE-14886 - removeTempOrDuplicateFiles is broken for list bucketing,
> +          // TODO: for IOW, we also need to count in base dir, if any
> +      // TODO: HiveFileFormatUtils.getPartitionDescFromPathRecursively for MM tables?
> +        // TODO: not having aliases for path usually means some bug. Should it give up?
>      updateMrWork(jobConf);  // TODO: refactor this in HIVE-6366
> +      // TODO: this assumes both paths are qualified; which they are, currently.
> +    // TODO: just like the move path, we only do one level of recursion.
> +  // TODO: may be broken? no LB bugs for now but if any are found.
> +  String lbDirSuffix = lbDirName.replace(partDirName, ""); // TODO: wtf?
> +        // TODO: we should really probably throw. Keep the existing logic for now.
> +    // TODO: this assumes both paths are qualified; which they are, currently.
> +        // TODO: we assume lbLevels is 0 here. Same as old code for non-MM.
> +          // TODO: we could just find directories with any MM directories inside?
>    // TODO: add test case and clean it up
> +  // Duplicates logic in TextMetaDataFormatter TODO: wtf?!!
> +  // TODO: why is this in text formatter? grrr
> +      // TODO# noop MoveWork to avoid q file changes in HIVE-14990. Remove (w/the flag) after merge.
> +    // TODO: this should never happen for mm tables.
> +    // TODO: if we are not dealing with concatenate DDL, we should not create a merge+move path
> +    // TODO: wtf? wtf?!! why is this in this method?
> +    // TODO: this relies a lot on having one file per bucket. No support for MM tables for now.
> +        // TODO: should this use getPartitionDescFromPathRecursively?
> +      // TODO: we should probably block all ACID tables here.
> +              @SuppressWarnings("unused") // TODO: wtf?
> +    // TODO: due to the master merge, tblDesc is no longer CreateTableDesc, but ImportTableDesc
> +        // TODO: ReplCopyTask is completely screwed. Need to support when it's not as screwed.
> +          // TODO: ReplCopyTask is completely screwed. Need to support when it's not as screwed.
>    // TODO: the below seems like they should just be combined into partitionDesc
> +    // TODO: could we instead get FS from path here and add normal files for every UGI?
> +-- TODO: doesn't work truncate table part_mm partition(key_mm=455);
> +-- TODO: need to include merge+union+DP, but it's broken for now
> +  // TODO: move these test parameters to more specific places... there's no need to have them here



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)