BatchJobs_1.9: - Maintenance release. Removed calls to deprecated `default.stringAsFactors()`. BatchJobs_1.8: - Maintenance release. Fixed some warnings. BatchJobs_1.7: - Configuration files are now sourced while loading the namespace, not while attaching the package. - Added OpenLava support - getJobInfo() now also displays time.running until an error occurred in the job, and not NA as before - loadRegisty() now returns a read-only registry if a heuristic detects that the location of the registy changed and the new argument "adjust.paths" is not explicitly set to TRUE. - Fixed a bug where time.started was incorrectly stored if chunking was used. - Packages are now loaded in user provided order instead of alphabetically. - showClusterStatus now returns its information visibly and does not print it per default - The 'max.load' setting for multicore and SSH is now set to a more aggressive default. - New argument 'impute.val' for loadResult() and loadResults() - New config options: "measure.mem" to disable memory measurements on the slave - SSH: makeSSHWorker, debugSSH: ssh.cmd and ssh.args were added to customize the remote login - New functions: -- setRegistryPackages() -- reduceResultsDataTable() -- getJobLocation() -- batchQuery() BatchJobs_1.6: - Added functions: getStatus, removeRegistry, findDisappeared - All find* function now support a 'limit' argument (default unlimited) - We are less strict in arg checks of makeSSHWorker, so you can set a load higher than ncpus - loadExports: added argument "what" - batchApply now works with arrays - filehandling for the registry, especially on Windows, was improved (Thx Henrik Bengtsson) - You can set pragmas via db.options in your config file. - The SQLite pragma "busy_timeout" is now set per default with a value of 5000ms. - The default number of cores is now retrieved from the option "mc.cores" rather - New configuration option BatchJobs.load.config - Fixed some minor things. BatchJobs_1.5: - Updated to use RSQLite >=1.0.0 - Fixed a bug in waitForJobs(). - Fixed a regression parsing array job ids. - Argument 'progressbar' added to multiple functions to suppress displaying the progress bars. - Configuration option 'staged.queries' is now enabled per default. BatchJobs_1.4: - Added functions: batchExport, batchUnexport, loadExports - Added functions: addRegistryPackages, removeRegistryPackages - Added functions: addRegistrySourceFiles, addRegistrySourceDirs, removeRegistrySourceFiles, removeRegistrySourceDirs - Added functions: getJobParamDf - New option: "BatchJobs.clear.function.env". If TRUE the function environment will be erased prior to writing it to the file system. Otherwise a warning is issued if the environment exceeds 10Mb in size. - version 1.3 introduced a mini bug in the SGE interface (via new list.jobs.cmd option). Fixed. - Memory usage is now stored in the database and can be queried using getJobInfo(). BatchJobs_1.3: - Reduced memory consumption and disk i/o if 'more.args' is used. - added option in Registry for absolute paths src.files and src.dirs - Added option "BatchJobs.verbose" for report generators like knitr - added "file.dir" option to batchMapQuick, removed option 'temporary' as this is now more flexible. - package 'methods' is always loaded on slaved to avoid problems with S4 code - some smaller fixes - in makeClusterFunctions* it is now to configurable via what command jobs are listed on the system - default for BatchJobs.check.posix is now: OS != Windows BatchJobs_1.2: - Fixed a bug where too many running jobs were queried on some Torque systems - Fixed a bug where chunking did not work with batchMapQuick - Fixed mail subjects to be more precise on job IDs - Fixed output of running time in testJob() - More robust error handling in syncRegistry - waitForJobs is now more conservative in waiting for termination of jobs. - New argument "impute.val" for reduceResults-family of functions: This allows to conveniently reduce results while some are missing - Experimental framework to support sourcing of multiple files and directories - testJob can now run jobs interactively in the current R session if argument "external" is set to TRUE. BatchJobs_1.1-1135: - Increased minor version number - Added support for SLURM schedulers - Function getErrors renamed to getErrorMessages due to conflicts with RUnit. Interface change: argument "print" dropped - Redefined return value of function "waitForJobs" which should now be much easier to understand. - New option "BatchJobs.check.posix" to enable more relaxed checks on file and directory names - Helper functions to simplify integration into other packages: getConfig, loadConfig and setConfig. showConf was replaced with a print generic. - Improved stability and speed of killJobs - Added a workaround to fix file URI handling on windows - Many minor bug and documentation fixes - The registry now supports an "exports" directory of exported objects, which are always loaded on the slaves, e.g. data objects which are needed in every job This directory can be either managed manually or - more conveniently - via the "fail" package on CRAN. - The registry now supports a list of source directories. All R files in these subdirectories are sourced on the slaves (and also on the master) Nice if you have a lot of helper code in seperate R files which is needed in your jobs. - batchMap and all result functions where this is reasonable support naming results with the names of the elements that were mapped over (see use.names argument) - batchMap now supports mapping over any kind of object (not only lists and vectors) that supports the "length" and "[" / "[["- index method. BatchJobs_1.0-966: - Dropped support for R < 2.13.0 (.find.packages deprecated) - speedups in reduceResults[ReturnValue] - Output of showStatus now contains a time information - fixed a bug in findErrors (argument 'ids' was ignored) - New function sweepRegistry: Remove intermediate/temporary/obsolete files - Better error handling in case of I/O errors - New functions: findStarted() and findNotStarted() BatchJobs_1.0-915: - new option to raise warnings to errors on the slaves - option to set process priority via nice in clusterFunctionsMulticore and clusterFunctionsSSH - new helper functions: getErrors, getJobInfo, loadConfig, callFunctionOnSSHWorkers, getSSHWorkersInfo, installPackagesOnSSHWorkers, waitForJobs, several more find* functions - removed helper getJobTimes -> this is included in getJobInfo - experimental file-based caching mechanism to speed up the processing of queries generated on computational nodes - various smaller fixes - documentation fixes BatchJobs_1.0-606: - many dependencies are now imports - many fixes and improvements for stability in cluster functions - helper functions for your own cluster functions are exported and documented, they all start with cf* - interactive mode now has dummy functions for listing and killing jobs - interactive mode now generates log files - job resources (from submits) are stored and can be queried on the master and the slaves (see getResources and getJobResources) - defaults for job resources can be specified in config file, see config part on web page - multicore and SSH mode now supports r.options, i.e. options to start Rscript and R CMD BATCH with - function grepLogs to search log files for a pattern BatchJobs_1.0-527: - very minor bug fixes: for argument conversion and checks BatchJobs_1.0-485: - First submit to CRAN.