> On Sept. 21, 2018, 2:11 p.m., kalyan kumar kalvagadda wrote: > > Arjun, > > > > Can you use an example to explain the optimization this chnage is going to > > make? > > Arjun Mishra wrote: > Kalyan, will give you an example. ReadEntity has an attribute called > accessedColumns. Say a table has 10 partitions and 7 columns. If we run a > select on that table Hive will build the ReadEntity inputs for us as > 1. 1 table input - Database.table. Will have 7 accessedColumns > 2. 10 partition inputs - Each partition input will have 7 accessedColumns > > Because of this we will build an inputHeirarchyList of size 77 (1*7 + > 10*7) like Server->Database->Table->Column. This inputHeirarchyList will have > duplicates. When sentry is checking for authorization it will be checking > over duplicates which is unnecessary
Making it a Set of List<DBModelAuthorizable> will eliminate chances of any duplicates making the authorization check much faster - Arjun ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/68787/#review208857 ----------------------------------------------------------- On Sept. 26, 2018, 5:11 a.m., Arjun Mishra wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/68787/ > ----------------------------------------------------------- > > (Updated Sept. 26, 2018, 5:11 a.m.) > > > Review request for sentry, kalyan kumar kalvagadda, Na Li, and Sergio Pena. > > > Bugs: SENTRY-2406 > https://issues.apache.org/jira/browse/SENTRY-2406 > > > Repository: sentry > > > Description > ------- > > When building input or output hierarchy list, we iterate over all ReadEntity > inputs received from Hive. The inputs particularly have accessed columns that > can repeat for other ReadEntity objects. This happens definitively when a > table has partitions. We should in general protect Sentry from not having to > authorize over a list of DBModelAuthorizable objects when it has already been > done. > > > Diffs > ----- > > > sentry-binding/sentry-binding-hive-common/src/main/java/org/apache/sentry/binding/hive/authz/HiveAuthzBinding.java > 6a1556fa7 > > sentry-binding/sentry-binding-hive/src/main/java/org/apache/sentry/binding/hive/authz/DefaultSentryValidator.java > f0764767f > > sentry-binding/sentry-binding-hive/src/main/java/org/apache/sentry/binding/hive/authz/HiveAuthzBindingHookBase.java > da1956b8e > > sentry-binding/sentry-binding-hive/src/main/java/org/apache/sentry/binding/metastore/MetastoreAuthzBinding.java > 5beda9f80 > > sentry-binding/sentry-binding-hive/src/main/java/org/apache/sentry/binding/metastore/MetastoreAuthzBindingBase.java > 0909656be > > sentry-binding/sentry-binding-hive/src/main/java/org/apache/sentry/binding/util/SentryAuthorizerUtil.java > dd6936c19 > > sentry-binding/sentry-binding-hive/src/test/java/org/apache/sentry/binding/hive/TestHiveAuthzBindingHookBase.java > PRE-CREATION > > sentry-binding/sentry-binding-hive/src/test/java/org/apache/sentry/binding/hive/TestHiveAuthzBindings.java > 3bbf6fb19 > > sentry-core/sentry-core-model-db/src/main/java/org/apache/sentry/core/model/db/AccessURI.java > dd8104d72 > > sentry-core/sentry-core-model-db/src/main/java/org/apache/sentry/core/model/db/Column.java > 305fd1f5a > > sentry-core/sentry-core-model-db/src/main/java/org/apache/sentry/core/model/db/Database.java > 0d94805c7 > > sentry-core/sentry-core-model-db/src/main/java/org/apache/sentry/core/model/db/Server.java > 33e735ef2 > > sentry-core/sentry-core-model-db/src/main/java/org/apache/sentry/core/model/db/Table.java > b161d0291 > > sentry-core/sentry-core-model-db/src/main/java/org/apache/sentry/core/model/db/View.java > f6d84997e > > > Diff: https://reviews.apache.org/r/68787/diff/3/ > > > Testing > ------- > > > Thanks, > > Arjun Mishra > >
