* Improve diffexp for tiledb
- The rows from the A and B sets are gathered and processed at the same time. In this
way the matrix is only accessed once instead of twice for each tile.
- There is now a single thread queue that gets shared between all callers of the diffexp.
This will slow down work if diffexp gets too busy.
- There is a target_workunit amount of work given to each thread. Previously the
workunit was (rows selected * width of tile), which could be small. Now multiple
column tiles can be combined into one workunit. If the target is too small then
thread and other overheads may reduce performance. If target_workunit is too large
then the size of the gathered sub matrix may take up too much memory.
- add configuration parameters (max_workers, cpu_multiplier, and target_workunit)
* Specialize diffexp for tiledb
This patch adds a new diffexp algorithm which is tuned for tiledb.
This algorithm was written by Bruce and is adapted here to plug into the
current framework. The anndata_adaptor still calls the original
algotithm (which was move from diffexp.py to diffexp_generic.py).
The cxg_adaptor now calls the new diffexp_tiledb version. Some
code is shared between the two.
This is part 1 of the diffexp for tiledb. Further tuning and
global throttles are still needed.
A script to run and time diffexp with various options is also
added: test/run_diffexp.py.
* s3 region should have a single config param
The s3 region can also now be automatically determined to further
reduce errors.
This patch also fixes a bug with order of handling the config params.
The tiledb config needs to be fixed before attempting to load
(need to handle_adaptor before handle_single_dataset)
* warning on maxCount for diffexp
* cleanup logging
* clarification
* make the limits configurable
* make diff exp limit work
* danger!
* remove debugging code
* fix merge with master
* fix unit tests
Co-authored-by: Colin Megill <colinmegill@gmail.com>
* Improvements to the matrix cache
- Add a timelimit for the matrix in the cache.
Once the timelimit is reached, the matrix can be removed.
- If a DatasetAccessError occurs, then remove the dataset
from the matrix cache.
Fixes#1322
* Added a config hook for secret key into the app.
the server first looks in an environment variable,
then looks in a config file.
For the cellxgene launch app, a default key is used if none is provided.
For the eb app, a secret key must be provided.
* Improved fix for matrix cache handling.
During the MatrixDataCacheItem acquire function there was a
time when the write lock was released and the read lock was taken.
During that time, the dataset could have been deleted, later
result in the MatrixDataCacheManageri data adaptor returning None.
The solution is to demote the writer lock to a reader lock instead
of unlocking and relocking.
Also, when a the cache needs to delete an entry, the delete
is done outside the MatrixDataCacheManager lock. This operation
only requires the write lock for the MatrixDataCacheItem.
Fixes#1255
* Introduce a config file to cellxgene
The config file format is in yaml. The default config is located
in server/common/default_config.py. A user may create a yaml file
that contains a subset of these fields. It can be used during cellxgene
launch, or for hosted cellxgene.
The code has also been refactored. Much of the logic to check arguments
has moved from launch to app config.
It is now possible to set the tiledb context parameters using the config
file. Other feature will soon be handled in a similar way.
