Changeset: b50fa802d315 for MonetDB
URL: https://dev.monetdb.org/hg/MonetDB?cmd=changeset;node=b50fa802d315
Modified Files:
        monetdb5/modules/mosaic/TODO_MOSAIC.txt
Branch: mosaic
Log Message:

Update TODO's.


diffs (26 lines):

diff --git a/monetdb5/modules/mosaic/TODO_MOSAIC.txt 
b/monetdb5/modules/mosaic/TODO_MOSAIC.txt
--- a/monetdb5/modules/mosaic/TODO_MOSAIC.txt
+++ b/monetdb5/modules/mosaic/TODO_MOSAIC.txt
@@ -1,19 +1,19 @@
 ####technical depth####
+Check if compressors are always called in tests. Use MOSdestroy trick for that.
 Make dictionary variable-sized so it can always compress. We are going to use 
vheap for this.
 dictionary encoding should use the ordered property of its dictionary in the 
algebraic operators
-decouple estimation completely from compression: there should be two runs: one 
for estimating the blocks and one to actual compress according to the 
estimations.
 Test null semantics
 mosaic.select (c,nil:any_1, nil:any_1,true,true,(true|false)) is not working 
correctly: should result in c == nil or c!=nil(anti).
 Use bit vectors more aggressively in mosaic_delta --- then resolve test 
mosaic_delta_timestamp --- and possibly other compression techniques as well.
 Somehow force mosaic to only use specified compressions even if inefficient. 
This poses a design problem for dictionary compression though.
 Be sure alignment is correct. Use asserts for this.
-fix layout call.
+Check MOSanalysis and its test.
+Fix/check/test MOSlayout.
 Add test for float/double
 delta compression: check if the hardcoded compression expression are correct. 
See also (2)
 Add asserts on type degeneration for bit => bte daytime => lng timestmap => 
lng date => int.
 Fix the task-factor array.
 See if we can re-introduce landmark stuff somehow.
-Fix/check/test MOSlayout.
 benchmark
 Do estimation before mosaic heap creation.That saves a potential heap extend.
 
_______________________________________________
checkin-list mailing list
checkin-list@monetdb.org
https://www.monetdb.org/mailman/listinfo/checkin-list

Reply via email to