Need simpler way of putting conditionals inline with %>% about magrittr HOT 8 CLOSED

Without thinking implementation yet, I am thinking something like the following could be an idea, and can be kept simple in implementation...

mtcars %>% Or(nrow, when = rowcount)

more generally,

x %>% Or(fun, when)

where fun is a function reference or a lambda, and when is a (single-valued) condition.

perhaps something like

Or <- function(value, fun, when) 
{
    if (when) fun(value) else value
}

from magrittr.

my comment above... wrong login...

from magrittr.

This sounds reasonable... I think that, as a name "Or" is too much like a logical OR. Maybe something like fif would be appropriate.

A couple more thoughts:

• I like the idea of allowing an else function.
• It makes more sense to me to have the condition first, before the function(s) to run.

fif <- function(x, cond, if_fun, else_fun = identity) {
  if(cond) if_fun(x)
  else else_fun(x)
}

# Some examples
mtcars %>% fif(TRUE, nrow)
mtcars %>% fif(FALSE, nrow)
mtcars %>% fif(TRUE, nrow, head)
mtcars %>% fif(FALSE, nrow, head)

from magrittr.

Right, "Or" is probably not the best -- I quite like when one can "read" the statements.
Another option could be

x %>% when(condition, otherwise = fun)

which can also be "read". Could also have the additional (optional/identity) argument

x %>% when(condition, do = fun1, otherwise = fun2)

although the way this is read is a little different than the first "sentence".

Just a thought.

You're suggestion is fine, although a little insider knowledge is needed to figure out what fif means...

from magrittr.

That makes sense.

I also realized that you'd probably want to capture the unevaluated function call, so that you could pass in additional arguments. Using your suggestion of function, it would be something like

x %>% when(cond, fun1(opt1, opt2))

So that when cond is TRUE, it would call fun1(x, opt1, opt2).

from magrittr.

Right, and we should consider whether the condition should be allowed to depend on x, e.g.

x %>% when(is.na %>% any, do = fun1(opt1, opt2), 
           otherwise = fun2(opt3, opt4))

I am thinking we could implement all this using the defer function we discussed previously.

Something like (this won't work directly, just to show the idea.)

when <- function(x, condition, do = identity, otherwise = NULL) 
{
    cond <- defer(condition)(x)
    if (cond) defer(do)(x) else defer(otherwise)(x)
}

This would be nice if condition depends on x but more annoying when it doesn't.
Maybe one could think of a syntax to distinguish.

from magrittr.

This is also my number one wanted feature. How about some syntax like this:

input %>%
  branch(is.data.fame() %>%
         as.matrix() %>% 
         ensure_symmetric() ) %>%
  graph.adjacency()

branch takes a bunch of chains, the first element of the chain is a logical expression. If all chains inside of branch are false, then nothing happens to the data, it skips the branch entirely.

Here is another example with a three-way branch:

input %>%
  branch(is.data.frame() %>% graph.data.frame(),
         is.matrix() %>% graph.adjacency(),
        ) %>%
  plot()

EDIT:

• inspired by metalsmith-branch: https://github.com/ericgj/metalsmith-branch
• of course it would be possible to have nested branches

from magrittr.

This at all interesting anymore, as you can do

mtcars %>% {if(rowcount) nrow(.) else .}

to take the initial example? If so, you're welcome to re-open...

from magrittr.