save and save.image, with two main differences. First, any functions which have been mtraced (see the Save and Save.pos know how to deal with lazy-loaded objects set up via mlazy. Save() is like save.image(). Save.pos(i) saves all objects from the ith position on the search list in the corresponding ".RData" file (or "all.rda" file for image-loading packages, or "*.rdb/*.rdx" for lazyloading packages). There is less flexibility in the arguments than for the system equivalents.Save()
Save.pos( pos, path, ascii=FALSE)..mypack if "mypack" is a maintained-package).saveSave and Save.pos, which is normally invisible to the user, but can be helpful if there is a failure during the save process (for example, if the system shuts down unexpectedly). The workspace image is first saved under a name such as "n.RData" (the name will be adapted to avoid clashes if necessary). Then, if and only if the new image file has a different checksum to the old ".RData" file, the old file will be deleted and the new one will be renamed ".RData"; otherwise, the new file will be deleted. This also means that the ".RData" file will not be updated at all if there have been no changes, which may save time when synchronizing file systems or backing up.
If a function f is saved while mtraced and then restored into a different R session, there will be an error message when you try to run f. Assuming that the mtrace(f,FALSE) or mtrace(f) if you really want to debug f. However, you can avoid the problem in the first place by always calling Save and Save.pos; this is faster than manually untracing and retracing, and avoids losing any breakpoints. If you use the cd system in mvbutils, you will rarely need to call Save.pos directly; cd, move and FF will do it for you.
Two categories of objects will not be saved by Save or Save.pos. The first category is anything named in options( dont.save); by default, this is ".packageName", ".SavedPlots", "last.warning", and ".Traceback")'; you might want to add ".Last.value". The second category is anything which looks like a maintained package, i.e. an environment whose name starts with ".." and which has attributes "name", "path", and "task.tree". A warning will be given if such objects are found. [From bitter experience, this is to prevent accidents on re-loading after careless mistakes such as ..mypack$newfun <- something; what you meant, of course, is ..mypack$newfun <<- something<="" code="">. Note that the accident will not cause any bad effects during the current R session, because environments are not duplicated; anything you do to the "copy" will also affect the "real" ..mypack. However, a mismatch will occur if the environment is accidentally saved and re-loaded; hence the check in Save.]
path is normally inferred from the path attribute of the pos workspace. If no such attribute can be found (e.g. if the attached workspace was a list object), you will be prompted. If path is a directory, the file will be called ".RData" if that file already exists, or R/all.rda if that exists, or R/*.rbd for lazy loads if that exists; and if none of these exist already, then the file will be called ".RData" after all. If you specify path, it must be a complete directory path or file path (i.e. it will not be interpreted relative to a path attribute).save, save.image, mtrace in package mlazySave() #
Save.pos( "package:mvbutils") # binary image of exported functions
Save.pos( 3, path="temp.Rdata") # path appended to attr( search()[3], "path")Run the code above in your browser using DataLab