Ensured that R script with unclosed brackets treated as incomplete R statement #7610
Conversation
|
@rdstern Please could you test? thanks |
There was a problem hiding this comment.
Great. This now works perfectly with that example. I tried it both from the script window and also within the Edit > Script dialogue.
Two questions:
a) Does it also work with square brackets [ ] and also curly brackets { } which could also be used?
b) Now, if we open a bracket and then don't close it (or a quote) then I assume we can type anything and it will just continue "eating" the code, until I close. In interactive mode this could go on a long time. If I suspect that something needs closing is there anything I can type that will close a command - and give an error - but enable me interactively to not go on forever? I hope the question is clear?
It works with curly brackets but not square brackets. As I mentioned, currently there is a simple set of heuristics that aim to cover the most common cases. There are many aspects of R that is does not cover (e.g. single/double square brackets, single quotes, binary operators (apart from '+'), function declarations and other key words where the '{' is on the next line).
It will continue adding the code to a buffer ('eating' the code) until the heuristics consider the statement closed. The user cannot see what's in the buffer, so there is a risk that they don't know how to close the statement. Currently, the only way to clear the buffer is to exit R-Instat. Options:
I prefer the first or last option. RStudio has a good approach. They have an editor and a console. The user can type commands directly into the console (or paste/ctrl-enter commands from the editor). The console executes the lines sequentially. If the statement is incomplete then it changes the console prompt from '>' to '+'. If the user presses 'escape', then it clears the buffer and resets the prompt to '>'.
We could consider replacing the script window with a console window, and then using RScript to cover the whole R syntax. R-Instat's console window should then be nearly as good as the one in RStudio (we may omit some of its advanced features). I estimate that this would take around 40 hours including the changes to the RScript library (this is a very rough initial estimate). What's your view? |
|
@lloyddewit after discussion with @volloholic we are excited that you should go for your "total solution". It fits so well with having a script button on each dialogue, with our "halfway" dialogues and so on. Very exciting. And just 40 hours too - I mean that's less than 2 days! That suggestion made my day yesterday, and the teaching in India, I was doing, has only confirmed that there will be powerful people who would welcome this facility. (The key person on that workshop, is a matlab user, who would quickly want to be using RStudio - i.e. commands, while having to start with R-Instat, because of its initial facilities in tidying and initial processing of the climatic data.) This also fits my strategy, that we can never be the easiest R GUI, but we can be easy enough, and (by far) the most powerful! And letting people start using R so easily, and then including the set of packages so they can proceed very quickly, is attractive in itself. |
|
@rdstern Thank you for your support and enthusiasm. We will go for it! |
|
I merged this PR and will implement the R console against issue #6744 |
Fixes #7609
In the script window, R statements can be split across multiple lines for readability. When R-Instat parses the script, it parses line by line, and only executes the text when it judges that the R statement is complete.
It uses a set of heuristics to judge if the statement is complete (e.g. unclosed quotes, line ends in ',' etc.). This PR adds a check to ensure there that all open brackets have been closed. See issue #7609 for an example.
Note: The set of heuristics is quite simple but should cover most real-world examples. The RScript library parsing is much more sophisticated. However, it doesn't currently have a public function to check if the R script represents an incomplete R statement.
I suggest to stay with the simpler heuristics for now. If we find that we continue to expand the heuristics, then we should consider updating RScript and using this library instead (see issue #6744).