search

KentonWhite / ProjectTemplate

A template utility for R projects that provides a skeletal project.
http://projecttemplate.net
GNU General Public License v3.0
623 stars 159 forks source link

New clear() function to manage global environment #167

Closed connectedblue closed 7 years ago

connectedblue commented 7 years ago

This PR proposes a new user function clear() to provide some useful tools for the ProjectTemplate user to manage their analysis workflow.

The motivation behind this is because some large datasets (e.g. raw or derived datasets) may not change all that often during analysis, but other variables do (e.g. models, frequency summaries etc). So a user may need to delete old variables and rerun (re)load.project() quite often but doesn't want to keep waiting for these large datasets to load from cache or data.

This function provides a new config item sticky_variables which can be configured with a comma separated list of variables that the user wants to keep in memory. So if the clear() command is issued, everything in global env will be deleted except those configured things.

The user can also specify ad-hoc things to keep without having to edit config/globals.dcf by using a parameter option: clear(keep=my_var).

If the user wants to really clear everything, they can issue a force: clear(force=TRUE).

As a bonus, there is a use case that works anywhere, not just ProjectTemplate: Typing clear() is so much easier than rm(list=ls()).

I've updated the website and manual documentation as well as including some tests for all the cases.

KentonWhite commented 7 years ago

Thank @connectedblue! Should there be a migrate for this new feature since a new config var is being added?

connectedblue commented 7 years ago

Hmmm ... good point @KentonWhite. Let me think about this. I'll play around with some scenarios. I tested that migration worked ok on my project, but I'm not sure if migration specific config is required.

I'll get back to you.

connectedblue commented 7 years ago

OK, I've checked what happens on when migrate.project() is run on on a previous project.

The following message is shown at the terminal:

Migrating project configuration
Your existing project configuration in globals.dcf does not contain up to
date configuration settings in this version 0.7 of ProjectTemplate.  They will
be added automatically during migration, but you should review afterward.
Your config.dcf file was missing entries and defaults
have been used.  The missing entries are:
 sticky_variables

This is what happens without any particular change for this file - the function picks up that there is a new config item automatically.

We don't need to set the default to anything different for migrated project, so I don't think there's a need for a custom migration behaviour.

KentonWhite commented 7 years ago

@connectedblue thanks for checking on that.