> On Feb. 5, 2016, 2:08 p.m., John Sirois wrote:
> > Can you note the technique you used for finding all last vestigaes of 
> > missing backfill in the testing done?
> > Was this grep-provable as the last bit to fix for example?
> 
> Maxim Khutornenko wrote:
>     That was mostly an oversight as we should have backfilled everything 
> touched in api.thrift: https://reviews.apache.org/r/42811/diff/3#1
> 
> Bill Farner wrote:
>     John - do you think this is something we should consider tackling with 
> codegen?  For example, if we could make the `IIdentity` and `ITaskConfig` 
> code do this it seems like we wouldn't be playing whack-a-mole.
> 
> John Sirois wrote:
>     I'm not sure what it would look like in codegen, though I'm not confident 
> enough in where the wrappers are used to be sure there still won't be a raw 
> unwrapped thrift struct used in a set somewhere.
>     I do think exploring wrapping the boundary interfaces - thrift rpc 
> ifaces, log storage, mappers - with an interceptor that has a map of type -> 
> backfill op though is worth exploring.  This is a handwave though since the 
> map would need to have a key for every thrift object that can contain a 
> thrift object needing backfill.  IE, X needs backfill, but we might have 
> `A.getB().getX()` so we'd need to backfill A's, B's and X's to get the X's.
>     
>     I assume we both agree that the better-than-whac-a-mole solution, if it 
> exists, can come in 0.13.0 though.
> 
> Bill Farner wrote:
>     > I assume we both agree that the better-than-whac-a-mole solution, if it 
> exists, can come in 0.13.0 though.
>     
>     Absolutely.

Here's what might be really useful in a make-believe world where my annotation 
work exists in Apache Thrift - annotating structs with key fields such that the 
generated objects are not value-equals, but keyed off a ubset of fields that 
form the logical key.  That I think would help alot with migrations.


- John


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/43262/#review118076
-----------------------------------------------------------


On Feb. 5, 2016, 2:03 p.m., Maxim Khutornenko wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/43262/
> -----------------------------------------------------------
> 
> (Updated Feb. 5, 2016, 2:03 p.m.)
> 
> 
> Review request for Aurora, John Sirois and Bill Farner.
> 
> 
> Bugs: AURORA-1610
>     https://issues.apache.org/jira/browse/AURORA-1610
> 
> 
> Repository: aurora
> 
> 
> Description
> -------
> 
> Backfilling JobConfiguration.Identity
> 
> 
> Diffs
> -----
> 
>   src/main/java/org/apache/aurora/scheduler/storage/log/ThriftBackfill.java 
> 9c88f5763d77ec0a304fdb2530ef6b6c1b0f230e 
>   src/test/java/org/apache/aurora/scheduler/storage/log/LogStorageTest.java 
> 13fcf9fd3afc14cf572d4bc7de355ff698b2c222 
>   
> src/test/java/org/apache/aurora/scheduler/storage/log/SnapshotStoreImplTest.java
>  3b176db12369bc5a2a63cb0018ffca7ef8b9055d 
>   
> src/test/java/org/apache/aurora/scheduler/storage/log/ThriftBackfillTest.java 
> 506d5919b9db14e04b10190a5f2d2aef123ac1dc 
> 
> Diff: https://reviews.apache.org/r/43262/diff/
> 
> 
> Testing
> -------
> 
> ./gradlew -Pq build
> 
> 
> Thanks,
> 
> Maxim Khutornenko
> 
>

Reply via email to