setup_reprex()

[moodymudskipper]

setup_reprex(fun, n = 1, nse = NULL)

• fun : a function
• n : number of reprexes to produce
• nse: vector of args not to construct,

This basically traces the function for n calls, it opens a new untitled script every time, and constructs arguments and the call.

This is not the most general way to approach things because NSE and environments are complex, but this should really be ok for 99% use cases I think.

Call without a fun or n arg from inside the function itself to create a reprex while debugging, using locally modified values if relevant

[moodymudskipper]

We have a naming issue because we have 3 candidate reprex related features :

• reprex on error error reprex #20
• construct to clipboard/viewer with reprex wrapper Construct to clipboard or into active document? #223
• construct reprex when entering a function (this)

That's why I named this setup_reprex but it's not a great name either

Maybe:

• construct_on_error()
• construct_to_reprex()
• construct_context()

[moodymudskipper]

The general way to deal with this is tricky, we should capture every argument's expression and environment (the env should be parent.frame() except for dots, we might use enquo() there), then check if the symbols used are accessible from that env, reconstruct these environments (at least the required objects, but even that is not completely general), then we we should use some quosures and inject() to be able to run the correct call, with BOTH same expressions and same values. This isa headache, I think we can have a "naive" mode and an "advanced" mode, and add an arg for this later, and just code the default "naive" approach for now, and if we're lucky we'll never really need the advanced one.

The draft PR #100 was trying some clever approaches but it's a can of worms.

[moodymudskipper]

Let's close for now