microsoft
/
rudder
зеркало из https://github.com/microsoft/rudder.git
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность
363 коммитов 17 Ветки 0 Теги 986 KiB
Граф коммитов

363 Коммитов

Автор SHA1 Сообщение Дата
Mike Barnett a0d7d8c1f0 Fixed line endings. 2017-12-14 11:03:21 -08:00
Mike Barnett 760f91a88b Avoid trying to create an instance of any processor which has a Finalize method. There is no way to guarantee that the method won't throw an exception and if it does, then the process is torn down because there's no way to catch it. 
Don't bother trying to load the core assembly greedily --- not sure why that was ever there.
Fix the caching of units so it actually can get used and avoid a possible bug by directly using the scopeRuntime assembly instead of searching all located units. The latter can cause an exception since the collection of loaded units can be modified.
 2017-08-01 13:28:39 -07:00
garbervetsky 63dc07d10a Remove call to (the incomplete) static produces method analyzer. 2017-07-19 11:58:43 -07:00
garbervetsky c801135900 Add support to mimic a copyTo when output columns are not written but there is a yield return current. 2017-07-19 11:10:02 -07:00
Mike Barnett 0e07bce77e Make a special log message for when we cannot use Reflection to get an instance of the processor. 2017-07-17 16:23:50 -07:00
garbervetsky 9ffff8424c Merge branch 'ConvertToCCI' of https://github.com/Microsoft/rudder into ConvertToCCI 2017-07-17 15:28:24 -07:00
garbervetsky 0ddf3f0058 Fix minor bug in Column handling 2017-07-17 15:27:43 -07:00
Mike Barnett f10851ce0b Make sure processor name is in output log. 2017-07-17 15:24:59 -07:00
garbervetsky cb02c8dac6 Solve minor conflict 2017-07-17 08:41:48 -07:00
garbervetsky 00e350b683 Just a minor refactor in BulkAnalysis 2017-07-17 08:41:10 -07:00
garbervetsky 5ae470d615 Merge branch 'ConvertToCCI' of https://github.com/Microsoft/rudder into ConvertToCCI 2017-07-14 17:41:19 -07:00
garbervetsky c362ea1df3 Add support for AddLast, AddFirst (actually add*) 2017-07-14 17:38:28 -07:00
Mike Barnett e5ba464fe9 Fix merge conflict. 2017-07-14 17:06:22 -07:00
Mike Barnett 33d779a03a Merge branch 'ConvertToCCI' of https://github.com/Microsoft/rudder into ConvertToCCI 
Conflicts:
	ScopeProgramAnalysis/Program.cs
 2017-07-14 17:03:51 -07:00
Mike Barnett b70fb99384 Track declared dependencies for each processor, but only if the user indicated that they allow column pruning. 2017-07-14 17:03:03 -07:00
garbervetsky aa57025760 Add analysis times computation 2017-07-14 16:41:37 -07:00
garbervetsky b11aa629f0 Just the initials bits to start the static analysis of Produces methods 2017-07-14 16:41:14 -07:00
garbervetsky 4524999bd3 Compute stats about analysis times 2017-07-14 16:39:29 -07:00
Mike Barnett f925ae8fb5 Add generated cross-apply processors to the list of processors that we do *not* analyze 
Change the output format for the declared passthrough columns to use '|' as a separator instead of "," to avoid confusion when opening the results with Excel.
Rename method.
Fix bug that caused a crash if the constant was a long: we cannot represent that as a RangeDomain because they are too long. Not important to lose that since they represent column indices.
 2017-07-14 13:21:26 -07:00
Mike Barnett 46020c426f reformat code 2017-07-14 10:37:08 -07:00
garbervetsky e02efde840 Fix a null pointer exception when a variable has no Traceable of type Table 2017-07-13 15:02:59 -07:00
Mike Barnett cc6b9edb3e Fix the runtime loader to be as simple as possible, also correct (it has to be an instance and not a set of static methods) 2017-07-13 13:49:12 -07:00
Mike Barnett bdf3496d3c Files missed on last commit. 2017-07-12 20:23:29 -07:00
Mike Barnett 1447f2766f Merge branch 'ConvertToCCI' of https://github.com/Microsoft/rudder into ConvertToCCI 2017-07-12 20:22:14 -07:00
Mike Barnett bc18e912c2 Move the loading of the Scope runtime into a separate project so it can be shared by both ScopeAnalyzer and ScopeProgramAnalysis. 2017-07-12 20:22:04 -07:00
garbervetsky ea4236dc3a Fix scripts to compute stats 2017-07-11 21:29:46 -07:00
garbervetsky a5539d204c change ',' delimiter for '+' to avoid problems with generics types with multiple parameters 2017-07-11 17:30:35 -07:00
Mike Barnett 28d0ca89e6 Sigh. Try to really fix it this time. 2017-07-10 16:19:02 -07:00
Mike Barnett 852e35b93f Fix null dereference bug. 2017-07-10 16:06:22 -07:00
Mike Barnett 066bb585d6 Make sure tool notifications get into dependency stats. 2017-07-10 16:03:07 -07:00
garbervetsky a13100b93f Remove subtration of extra analysis overhead to get baseline results. 2017-05-12 17:47:20 -03:00
Mike Barnett f1cc3cd24e Always have ExtractDependencyStats retrieve the schema inputs and outputs, even if the dependency analysis returns Top. 2017-05-12 10:33:43 -07:00
garbervetsky ff446444c5 Added one column to see if Zvonimir analysis returned top 2017-05-12 00:40:36 -03:00
garbervetsky f6878becfc Modify a little bit the output to compare Zvonimir analysis with dependency analysis 2017-05-12 00:02:24 -03:00
garbervetsky 1eb8d3ddbc Merge branch 'ConvertToCCI' of https://github.com/Microsoft/rudder.git 2017-05-11 21:16:47 -03:00
Mike Barnett b9aae8d637 More changes to stats. 2017-05-11 17:08:02 -07:00
garbervetsky e4bea8d7b9 Merge branch 'ConvertToCCI' of https://github.com/Microsoft/rudder.git 2017-05-11 20:02:47 -03:00
garbervetsky cc654537fe forget to save a file 2017-05-11 20:02:41 -03:00
garbervetsky ee08f27d91 Partial implementation of stopwtach for extra analysis overheads 2017-05-11 20:02:01 -03:00
Mike Barnett 3d556a7f7a Trying to get the statistics that are printed in better shape. 2017-05-11 15:52:26 -07:00
Mike Barnett 17cf72d6c3 lots o' stuff 2017-05-11 15:07:57 -07:00
garbervetsky fa12555fab Premilimary support for ScopeMaps (very rough yet) 2017-01-19 16:19:18 -03:00
Mike Barnett b4fac67fe9 Add a processor that conditionally reads from either column A or column B depending on whether A exists in the input schema. 2017-01-06 11:57:49 -08:00
Mike Barnett 2bcefc41ef Make the code that creates an instance of the processor and calls its Produces method to get passthrough information more robust. 2017-01-06 10:55:38 -08:00
Mike Barnett 25a9bff4ac Added unit tests for processors that just return a prefix of the input table. 2016-12-22 13:46:28 -08:00
Mike Barnett 9ae9fb6a1f Add a dependency on the real runtime and use reflection to find out what passthrough columns are actually declared by the processor. 2016-12-22 13:38:31 -08:00
Mike Barnett bfa8fc4420 Add the type "ScopeMap" to the runtime. 
Add a reducer that uses a ScopeMap (and performs "column promotion" with it).
Add a unit test for the reducer.
 2016-12-20 10:15:37 -08:00
Mike Barnett c7ba5d658c Fix bug where a processor might be a "ConcurrentProcessor", which is a generic class that takes as its first type argument the real processor. 2016-11-07 10:30:17 -08:00
Mike Barnett 51eb503404 Change log and dependency stats so the two analyses are not directly compared (since each can be more conservative than the other in different circumstances), but at least it is logged whether each analysis returned "top". 2016-11-04 10:14:39 -07:00
Mike Barnett c37dcc79c0 Throw an exception if the ScopeRuntime or ScopeVertexDef.xml cannot be found. 
Make the dependency statistic extraction (from the log) eager instead of lazy. No real reason for it to be lazy and having it lazy made it harder to catch exceptions from it.
Make the code a bit better for searching for the runtime.
 2016-11-02 15:04:11 -07:00