Bumped version number to reflect changes in basejump dependencies.
import: Bug fix for invalid objects (e.g. S4 objects that inherit
SummarizedExperiment) not returning silently. Had to convert the
trycall to a
tryCatchcall to avoid errors popping up during name checks.
import: Now suppressing
partial match of 'OS' to 'OS.type'warning for import of XLS files, which is a bug in gdata package.
importdocumentation, for clarity.
data.table::freadfor import of bcbio count matrix. Previously, was using
import: Don’t attempt to slot attributes for atomic vector return. This applies to source code lines and helps avoid valid name issues when assigning these values to colnames or rownames. I came across this issue while updating the Chromium package to assign names from 10X Genomics sidecar files.
export: Reworked internal code to call
data.table::fwritedirectly, rather than having to pass to
brio::export. Added support for
import: Improved internal arguments passed to
import: Improved metadata stash approach inside S4 (
metadata) and S3 (
attributes) return objects. Previously, import metadata was stashed inside “brio”, but this has been renamed to “import”. Metadata is no longer stashed inside R data objects loaded via
import. Simplified the internal code inside
importto only stash the
call, whereas importer metadata is now handled by an internal
Added back internal support for readr package instead of data.table for
export functions. We have observed stack imbalance and segfault memory dump issues with the latest data.table release (v1.12.4) on multi-core Azure VMs. The engine can now be changed using global options:
acid.import.engine(“data.table” or “readr”).
acid.export.engine(“data.table” or “readr”).
This new addition is experimental and may be dropped in a future release. We find that readr currently works more reliably for export in some cases for large CSV files, but data.table is generally faster and more robust for data import of CSV and TSV files. We’re intentionally keeping this functions simple and not providing a user-facing argument for selecting the internal engine.
naStrings: Reverted back to including only “NA” and “NULL”, instead of including empty space strings. This results in unwanted messages regarding strip whitespace from data.table
getURLDirList: Return a simple character vector of files and subdirectories in a remote directory. Intended for use with FTP servers.
getURLfunction from RCurl.
export: Removed internal dependencies on
as.data.tablecalls. The tibble package recently changed the default row name handling behavior in
as_tibble, which broke the code here. I reworked the internal code to only use base R approaches, so changes in the tidyverse no longer affect the package.
metadataparameter option and improved internal
tryCatchhandling if call capture fails. This can be the case when nesting the function inside another function, which can cause
standardizeCallto fail. Note that
match.calldoesn’t have this problem but doesn’t consistently expand the call with default formals as well.