Tailor-made functions and operations

This vignette presents how to customize stringmagic to (better) suit your needs. It covers:

Creating new functions with alias generators

Several stringmagic functions dispose of alias generators, which end with the suffix _alias. They generate a copy of the function with different default values (and also take care of setting up the environment correctly). Some functions have many arguments, by changing the default arguments you can create completely new functions.

This section will guide you through alias generation, and how they can be useful, using examples.

Creating a formula builder

This is a common R problem: how to turn a character string into a formula, in a handy way? You can use a few arguments of string_magic to make it work:

  • .default: to apply a default sequence of operations to interpolations
  • .post: to appy a custom function right before returning the object

The objective here is to inject variable names into a character vector and turn it into a formula. Since variables in a formula are separated with a "+", we need to collapse several variables with "+". This will be achieved with the .default argument. Then we turn it into a formula by passing as.formula in the .post argument.

y = "Petal.Length"
x = c("Sepal.Length", "Petal.Width", "Species")
string_magic("{y} ~ {x}", .default = "' + 'collapse", .post = as.formula)
#> Petal.Length ~ Sepal.Length + Petal.Width + Species
#> <environment: 0x5572f45564a8>

Now that we see that our builder works with string_magic, we create a dedicated function with an alias.

fml_builder = string_magic_alias(.default = "' + 'collapse", .post = as.formula)

The function fml_builder works as string_magic but with different default values. Now we can apply it directly.

fml_builder("{y} ~ {x}")
#> Petal.Length ~ Sepal.Length + Petal.Width + Species
#> <environment: 0x5572f4b69640>

Since this function is just a call to string_magic, you can apply anything you want. Let’s scale the variables on the right-hand-side by using nesting:

x = c("Sepal.Length", "Petal.Width")
fml_builder("{y} ~ {'+'collapse ! scale({x})}")
#> Petal.Length ~ scale(Sepal.Length) + scale(Petal.Width)
#> <environment: 0x5572f0dbfdc8>

Changing str_clean

The function str_clean specialses in cleaning character vectors. To do so it uses a specific syntax to transform various regular expressions at once. For example "pat1, pat2 => replacement" will turn the regular expression pat1 and pat2 into the replacement. The syntax is: i) a comma separated list of regular expressions, 2) a pipe (" => "), 3) the replacement. If you are not happy with the fact that regexes are separated with commas, or of the look of the pipe, no problem!

Let’s change the default values, let’s: i) use a semi-colon separation, ii) use ">>" instead of the regular pipe.

my_clean = string_clean_alias(split = "; ", pipe = " >> ")

x = "My name is Bond, James Bond"
# old way
string_clean(x, "e, o => a")
#> [1] "My nama is Band, Jamas Band"

# new way
my_clean(x, "e; o >> a")
#> [1] "My nama is Band, Jamas Band"

Creating small numeric matrices

The function string_vec facilitates the creation of small character vectors. With its arguments .nmat, you can turn the vector into a numeric matrix. Hence, let us write a function to create small numeric matrices.

We will use:

  • .nmat = TRUE to ask to transform the result into a numeric matrix. The number of rows will be deduced from the number of newlines.
  • .last = "'[\n ,]+'split" (.last means last operation) to split the resulting character vector with respect to newlines, commas and space, so that numbers can be separated by any succession of these
num_mat = string_vec_alias(.nmat = TRUE, .last = "'[\n ,]+'split")
num_mat("1, 2, 3
         7, 5, 0
         0, 0, 1")
#>      [,1] [,2] [,3]
#> [1,]    1    2    3
#> [2,]    7    5    0
#> [3,]    0    0    1

Creating your own string operations

You can add any arbitraty operation to string_magic. There are two main ways:

  • registering a custom sequence of regular string_magic operations (string_magic_register_ops)
  • registering a custom function (string_magic_register_fun)

New operations as a sequence of existing operations

Create new sequence of operations with string_magic_register_ops. It takes two arguments:

  1. the sequence of string_magic operations,
  2. the name of the new operation.

For example, let’s create a new operation h1 which formats a string into a header. It adds an hypen before the text and adds hyphens after the text up to the 40th column.

Ex.1: creating a header operation.

# 1) we register the sequence of regular string_magic operations
string_magic_register_ops("'- | 'paste, '40|-'fill", "h1")
# 2) we use it
string_magic("h1 ! That's my header", .nest = TRUE)
#> [1] "- That's my header ---------------------"

New operations using a custom function

To implement new operations using functions, you have two steps: 1. create the custom function, 2. register the function an provide an alias to the operation.

First you need to create a function that will be applied to a character vector.

That function must have at least the arguments x and ....

Additionnaly, it can have the optional arguments: argument, options, group, group_flag. This function must return a vector. Optionnally, and only if relevant (see the last example), you can add an attribute "group" to the returned object which will be used in grouped operations.

Second, you need to register the function with string_magic_register_fun and assign an alias to it. Optionnally, you can provide a list of valid options.

Let’s create an example in which we add markdown emphasis to words.

Ex.1: a new operation adding markdown emphasis.

library(stringmagic)
# A) define the function
fun_emph = function(x, ...) paste0("*", x, "*")
 
# B) register it
string_magic_register_fun(fun_emph, "emph")

# C) use it
x = string_vec("right, now")
string_magic("Take heed, {emph, collapse ? x}.")
#> [1] "Take heed, *right* *now*."

More generally, the function taken by string_magic_register_fun is called internally by string_magic in the form fun(x, argument, options, group, group_flag). Here is the meaning of the arguments:

  • x: the value to which the operation applies.
  • argument: the quoted string_magic argument (always character).
  • options: a character vector of string_magic options.
  • group: an index of the group to which belongs each observation (integer).
  • group_flag: value between 0 and 2; 0: no grouping operation requested; 1: keep track of groups; 2: apply grouping.

The two last arguments, group and group_flag, are of use only in group-wise operations only if fun changes the length or the order of vectors.

Let’s add an argument and an option to the "emph" operation that we defined in Ex.1.

Ex.2: new operation with argument and option.

fun_emph = function(x, argument, options, ...){
  arg = argument
  if(nchar(arg) == 0) arg = "*"
  
  if("strong" %in% options){
    arg = paste0(rep(arg, 3), collapse = "")
  }
  
  paste0(arg, x, arg)
}

string_magic_register_fun(fun_emph, "emph", "strong")

x = string_vec("right, now")
string_magic("Take heed, {'_'emph.s, c? x}.")
#> [1] "Take heed, ___right___ ___now___."

# In string_magic_register_fun, the valid_option argument is used to validate them.
try(string_magic("Take heed, {'_'emph.aaa, c? x}."))
#> Error : in string_magic("Take heed, {'_'emph.aaa, c? x}."): 
#> CONTEXT: Problem found in "Take heed, {'_'emph.aaa, c? x}.",
#>          when dealing with the interpolation `{'_'emph.aaa, c? x}`. See
#>          error below:
#> The option `aaa` is not valid for the current operation.
#> FYI the option available is `strong`.

Finally let’s illustrate an example with group-wise awareness. This is somewhat advanced and should be of concern only when you regularly use group-wise operations.

Ex.3: we create a function that only keeps variable names (ex: x5, is_num, etc).


keep_varnames = function(x, group, group_flag, ...){
  is_ok = grepl("^[[:alpha:].][[:alnum:]._]*$", x)
  
  if(group_flag != 0){
    group = group[is_ok]
    # recreating the index
    group = unclass(as.factor(group))
  } 
  
  res = x[is_ok]
  # we add the group in an attribute (this is the way)
  attr(res, "group") = group
  
  return(res)
}

string_magic_register_fun(keep_varnames, "keepvar")

expr = c("x1 + 52", "73 %% 5 == x", "y[y > .z_5]")
string_magic("All vars: {'[^[:alnum:]_.]+'split, keepvar, unik, enum.bq ? expr}.")
#> [1] "All vars: `x1`, `x`, `y` and `.z_5`."

# thanks to the group flag, we can apply group-wise operations
# we apply cat after the function (using .post) to have a nice display of the newlines 
string_magic("Vars in each expr:\n",
             "{'\n'c ! - {1:3}) {'[^[:alnum:]_.]+'split, ",
                                 "keepvar, ~(unik, enum.bq) ? expr}}", .post = cat)
#> Vars in each expr:
#> - 1) `x1`, `x`, `y` and `.z_5`
#> - 2) `x1`, `x`, `y` and `.z_5`
#> - 3) `x1`, `x`, `y` and `.z_5`
#> NULL

Using stringmagic with custom operations as a dependency

This is section is only relevant if:

  1. you use stringmagic as a dependency in your package, and
  2. you use custom stringmagic operations.

If you answer “no” to any of these two points, do not read this section, it’s only about details. Otherwise it’s a must read.

To use custom stringmagic operations within a package, you need to explicitly register the operations in a specific namespace, and, when using the stringmagic functions, you need to add the argument .namespace telling where the new operations are located.

Why do I need a namespace?

There are two reasons: i) to ensure compatibility with future versions of the stringmagic package, and ii) to avoid conflicts with user-created operations. Let’s take an example. You create a package an use stringmagic with the custom operation h1, detailed here, to create headers. Now let’s say a future version of stringmagic also introduces a h1 function, that works differently from yours. This means that your package will work with the old stringmagic version but will lead to a bug with the new version. Not great!

Same story if the user defines her own h1 operation: we end up with two operations with the same name, hence a conflict.

We need a mechanism to ensure that the h1 operation in your package always work, irrespecive of the doings of the user and of the version of stringmagic. We now detail how to do it.

Using custom operations in a package

To use custom operations in a package:

  1. add the argument .namespace = "my_package_name" to the calls to string_magic_register_ops and string_magic_register_fun
  2. add the argument .namespace = "my_package_name" to any call to stringmagic’s functions using custom operation. Achive this simply by creating aliases to stringmagic function.

Example

You develop the package superpack which uses stringmagic to display messages to the user and you want to register the header operation (behaving similarly to h1).

string_magic_register_ops("'- | 'paste, '70|-'fill", 
                          alias = "header", 
                          namespace = "superpack")

You can now summon your new operation to write a message to the user. Just remenber that you need the .namespace argument:

time = 0.7
cat_magic("{header!Important message to you, user}",
          "The algorithm converged in {time}s.", 
          .sep = "\n", .namespace = "superpack")
#> - Important message to you, user -------------------------------------
#> The algorithm converged in 0.7s.

Without the namespace argument, this leads to an error:

time = 0.7
cat_magic("{header!Important message to you, user}",
          "The algorithm converged in {time}s.", 
          .sep = "\n")
#> Error: in cat_magic("{header!Important message to you, user...: 
#> CONTEXT: Problem found in "{header!Important message to you, user}\nThe
#> algorithm converged in {time}s.",
#>          when dealing with the interpolation `{header!Important message
#>          to you, user}`.
#> PROBLEM: `header` is not a valid operator. Maybe you meant `head`?
#> 
#> INFO: Type string_magic(.help = "regex") or string_magic(.help = TRUE)
#> for help.
#> Or look at the vignette:
#> https://lrberge.github.io/stringmagic/articles/guide_string_magic.html

Using the .namsepace argument is cumbersome. That is why stringmagic offers alias generators to easily create aliases of the stringmagic functions using custom operations.

Using aliases

To avoid providing the argument .namespace at each call (which makes the function completely useless), use the alias genetaors, as described in the first section. Let’s continue on the previous example but this time we avoid the argument .namespace by creating an alias.

To create the alias, we use cat_magic_alias which creates a copy of cat_magic for which the default values have been modified. Here we override the function name (of course you can use any other name, like cmagic, magcat, whatever!):

cat_magic = stringmagic::cat_magic_alias(.namespace = "superpack")

An now we are able to access our previously defined function without error:

time = 0.7
cat_magic("{header!Important message to you, user}",
          "The algorithm converged in {time}s.", 
          .sep = "\n")
#> - Important message to you, user -------------------------------------
#> The algorithm converged in 0.7s.