NEWS.md
Changes:
Explicitly settings DuckDB’s temp_directory to the Andromeda temp folder to avoid running out of space in an undefined temp location.
Set default memory limit to 20% instead of 80% of system memory to avoid out-of-memory errors when there are multiple Andromeda objects in memory.
flushAndromeda() now also evicts the cache by default, and is called after appending or creating a table from another Andromeda table.
The length() and names() method no longer throw an error when the Andromeda object is no longer valid, but just return 0 and an empty vector, respectively. This is to avoid error messages caused by RStudio querying the object.
Bugfixes:
Changes:
Leveraging DuckDB for faster assigns and copy operations.
Added flushAndromeda() function.
Added andromedaThreads option to control the maximum number of threads Andromeda is allowed to use.
Bugfixes:
Calling flushAndromeda() before copying entire Andromeda (copyAndromeda()) or just a table ([[]]<- operator) to avoid segfault.
Switch from zip::unzip() to utils::unzip() to avoid ‘mtimes’ errors.
Changes
Switch backend from SQLite to DuckDb for greater performance in terms of speed and disk space.
Added andromedaMemoryLimit option to control the maximum amount of memory Andromeda is allowed to use (in GB).
Bugfixes
isAndromedaTable() when table belongs to descendant of Andromeda.Bugfixes
arrow is not installed. Required to stay in CRAN.Bugfixes
Changes
names and names<- generic functions.names and names<- generic functions.Changes
Bug Fixes
Changes
createIndex, listIndices, and removeIndex functions.Bug fixes
Setting SQLite temp_store_directory to Andromeda temp folder as well to prevent running out of space in default drive.
Normalizing path before querying available disk space to avoid errors when using tilde in the path.
Bug fixes
Changes
Throw more informative error when user tries to use invalid Andromeda object.
Added function getAndromedaTempDiskSpace. Requires rJava to be installed. (Returns NA if not installed.)
Throw warning when disk space becomes low. Threshold defaults to 10GB, but can be altered using options(warnDiskSpaceThreshold = <n>).
Allow assigning query result to an Andromeda table where the query uses that Andromeda table. (e.g. andromeda$cars <- andromeda$cars %>% filter(speed > 10))
Added progressBar arguments to batchApply and groupApply.
Bug fixes
Unzipping now also done in Andromeda temp folder (instead of system temp folder).
Lower but faster compression level now also used when saving without maintaining connection.
Changes
Dropping nrow and ncol support, as this seems to cause instability and is not consistent with dplyr.
Dropping isSorted function, as database queries are only guaranteed to stay sorted if explicitly required to (using arrange).
Added restoreDate and restorePosixct functions.
Bug fixes
Fixed typo in isAndromeda function name.
Fixed incompatibility issue with dplyr 1.0.0 causing ‘method not supported’ error.
Correctly handling differences in column order when appending.
Fixed copying of zero-row tables from one Andromeda to another.
Fixed issue where Andromeda object inheritance was lost on R restart.
Changes
Minor edits to documentation in preparation for submission to CRAN.
Changing compression level when saving: 10-fold reduction in compression time at the cost of 10% larger file size.
Increasing batch size from 10,000 to 100,000 to increase speed.
Turning of SQLIte journal to increase speed.