Improve parsing of multi-line R script

[lloyddewit]

PR #6733 processes multiple lines of code in the 'Try' window. This PR works well for most cases. However, more sophisticated R parsing is available in the RScript library. I plan to create a NuGet package from this library in the coming months, and add it to R-Instat. I should then:

• Review the changes in PR Enabled New Data Frame dialog to 'try' running multiple lines #6733 again, and see if the parsing can be improved with the RScript library
• See if the Script window, or other dialogs, could be improved by using the RScript library

[rdstern]

That sounds great.

[lloyddewit]

I expanded this issue to include the proposal in PR #7610 :

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).

[rdstern]

@lloyddewit this is an exciting advance. When you made the recent change (to include brackets in a multiline command) I also checked that this change also worked in the Edit > Script dialogue. I am assuming the new console may make this dialogue obsolete, but I wonder if it may still have a role, because the scripts needed will often have a particular structure, that is made easy through this dialogue, namely:
a) Using particular packages
b) Getting data from an R-Instat data book
c) Doing something
d) Saving results back into the same data book, either adding to the data sheet, or maybe adding another data sheet into the data book
e) Tidying up - mainly deleting unneeded objects.

[lloyddewit]

@rdstern
RStudio has both a console window and an editor window. I think we need the same.

Initially, we can leave the R-Instat Edit > Script dialog functionality unchanged. If the user clicks the 'To Script' button, then the contents of the editor shall be sent to the new console window. The console shall then process the script just as if it was typed in line by line by the user.

In the future, we could enhance the Edit > Script dialog to allow single lines or selected text to be sent to the console.

I suggest that we look together at how the RStudio editor and console interact and clarify if/how we should implement this in R-Instat.

Do you have any time tomorrow morning (or next week)?

Thanks

[rdstern]

@lloyddewit Here is a script - well a function from the DescTools package documentation, function Fibonacci. It gives an error.

# Fibonacci numbers without iteration
fibo <- function(n) {
phi <- (sqrt(5) + 1)/2
fib <- (phi^(n+1) - (1-phi)^(n+1)) / (2*phi - 1)
round(fib)
}
fibo(30:33)

When I put it into the script window - development version - the function part. It gives an error and says:

[rdstern]

While I am thinking about it, I list some examples of sample scripts that could be well documented and available, for teaching purposes, and to facilitate tasks that are not so simple directly from the dialogues. These are just to record them and this does not affect the timing/priority of these improvements in R-Instat.
a) We have a tweaking guide - and tweaking dialogues, i.e. starting with a dialogue and then editing it slightly, will remain a key area. This will give examples where the starting point is a dialogue (or more dialogues), rather than a sample script file.
b) An example from there is the skew boxplot. Of course we might have a complete script for those who prefer to start from a script.
c) The example from the Text-Mining with R book, by Silge. This is neat R to read the 6 Jane Austen books and add information, including chapter headings for each book. It is an example to automate the tidying of the data.
d) Simple example of random sampling using the wakefield package. I give an example in the File > New Data Frame. There is a keyboard in the calculator, which works well, but I don't see how to use it easily to give variables containing missing values.
e) Examples that add "looping" when needed. I hope we will soon be able to do some of this with our new select objects in R-Instat dialogues, but it should be simple to repeat operations on similar variables, e.g. labelling factors, or columns generally, or analyses, using scripts designed for that purpose.
f) Examples of "topping and tailing". Taking a general script and adding the way of getting data from a data book, and writing results/data back into a data book. This links to one of my jobs, namely to document the R-Instat functions that do that.
g) Crop simulation modelling in R-Instat. This is being done by Sophie using RStudio. An example of the way this can then be done easily in an R-Instat script.
h) Same for our work on processing climatic data. Possibly adding homogen could be an example, but using Danny's scripts in R-Instat will be very useful.
i) More to come!

[lloyddewit]

Note:

PR #8551:

• Runs the Fibonacci numbers script above correctlt
• Resolves the tasks in the opening comment above

[rdstern]

Great! I am also reminded of this function, which I like, because we use it in the calculator practical.