TODO.md

In jmw86069/jamsession: Store and recall R sessions, objects, functions using a central location

jamsession TODO

06jun2022

Consider grep_functions() to be consistent with grep_jamsessions() and grep_objects().

28oct2021 - R objects

The save_object() mechanism should allow saving multiple objects, similar to saving an R session with a corresponding "objectlist.txt" file with the objects saved.

New idea: optionally check objects for size before saving session? If any objects are greater than 500MB it could exit with a warning.

• max_objectsize=500*1024*1024

22oct2021 - R sessions

General idea is to store the list of R objects in its own file to help search for R objects within a stored R session.

• save_jamsession() - argument save_objectlist=TRUE enables saving the list of Robjects using file extension defined by objectlist_suffix=".objectlist.txt".

05oct2021

• save_jamsession()

• should allow underscore and not convert to hyphen/dash: "-".

• some method to decide whether to save environment objects, default TRUE currently.

• prune_jamsessions() - utility to delete old versions of jam session files.

• prune_objects() - utility to delete old versions of R object files.

• grep_functions() - utility to search functions_path and list available files. Bonus points for listing functions defined? Probably easier to import.

• refresh_functions_for_package() - same as refresh_functions() except it creates a folder with proper package structure that can be used in an R package.

Generally need a better mechanism to store a few R objects in a file. Currently the R object names are concatenated to produce the output filename.

usability

• save_jamsession() objects to exclude or include

• exclude object names by pattern and/or vector of object names

• include only a defined vector of object names

  • could be good for saving a minimal R session of "reference data" for a given analysis project

• exclude object classes (e.g. environments)

  • for example: loaded a couple sessions into their own sub-environments. When saving the analysis session, the sub-environments probably do not need to be saved, preventing writing a bunch of un-necessary data to the RData file.
  • Bonus points for detecting and ignoring environments.

jamsession functions

• check virtual package name for direct conflicts with names of fully installed R packages

• warn/stop if virtual package would conflict and thus override an installed R package

• New functions to round out the workflow:

• list_functions() - list all stored functions

• grep_functions() - grep list of functions for keywords

deeper ideas into usefulness

• consider associating notes to an R session?

• "useful R script(s) associated with this session"

• Is it possible to grep stored functions for the name of a specific function?

• grep_function_name() - search by function name

  • option to subset the files searched by providing grep string(s).

• easiest method: Parse the .R file for specific pattern

  • This step makes assumptions that stored functions would follow a certain fixed pattern. However it should be fast across numerous files without having to source/load each file.
  • "^[a-zA-Z0-9_]+[ ]*<-[ ]*function" This pattern matches all function-compatible names starting on a new line that are immediately followed by <- function on the same line.

• more involved: load all functions .R files as temporary packages, this process would be too slow for routine searches across all .R files.

• most involved: process to "index" functions .R files:

  • Store the function details alongside each .R file, perhaps by running jamba::jargs() or storing formals() for each function.
  • In order to learn function details, each .R file would need to be loaded as a virtual R package in order to inspect the contents. This alternative runs this process only once per .R file.

  • Create a new index, or update an existing index when:

    • no index entry exists for that .R file
    • the index entry is older than the .R file
    • index entries can be stored in flat files, or central SQLite db

jamsession objects

• Document the use of object_notes_list argument.

• Consider returning object notes with load_object() as a proper R object for review.

• Mechanism to "describe" objects in an objects RData file. If someone searches for R-objects RData files by name, how would they know what's inside?

• currently when multiple objects are stored in one RData file, the filename is the set of objects separated by hyphen "-".

• List RData contents, run jamba::sdim() to describe size and class of all objects
• This step probably involves loading objects into a temporary environment then running jamba::sdim() on that environment

• This step will be slow for very large RData files, but it will work.

• Possible to save jamba::sdim() in a separate index, to prevent having to reload the full RData file to read its contents.

• Is there any R method to read contents of an RData file without loading/parsing the entire RData file?

  • I have seen this type of function in several custom R packages, they all seem to load the full RData file.
  • On StackOverflow help threads, the only solution I recall seeing was to load the full RData file into a temporary environment then inspect contents there.

jmw86069/jamsession documentation built on June 8, 2022, 8:47 p.m.