Current section
Files
Jump to
Current section
Files
lib/wild.ex
defmodule Wild do
@moduledoc """
Provides the interface to underlying wildcard implementations via the public
`match/2` and `match/3` functions.
"""
alias Wild.{Bash, Byte, Codepoint}
require Logger
@doc """
Executes a unix-style wildcard pattern match on a string with a given
pattern. By default it tokenizes and runs on Codepoints but can also be set
to Byte mode.
It supports all of the usual wildcard pattern mechanisms:
- `*` matches none or many tokens
- `?` matches exactly one token
- `[abc]` matches a set of tokens
- `[a-z]` matches a range of tokens
- `[!...]` matches anything but a set of tokens
## Examples
iex> Wild.match?("foobar", "foo*")
true
iex> Wild.match?("foobar", "fo[a-z]bar")
true
iex> Wild.match?(<<9, 97, 98>>, "?ab")
true
iex> Wild.match?("foobar", "bar*")
false
iex> Wild.match?(<<16, 196, 130, 4>>, "????", mode: :byte)
true
The options are:
* `:mode` - The matching mode. This primarily affects tokenization and
what is considered a single match for the `?` wildcard. Options are:
* `:codepoint` (default) - Tokenize on printable String characters
* `:byte` - Tokenize on each byte
* `:bash` - Using an underlying bash script. Only for debugging
The distinction is important for subject and patterns like the following,
where the binary is represented by two bytes but only one codepoint:
```
iex> Wild.match?("ā", "[!abc]", mode: :codepoint)
true
iex> Wild.match?("ā", "[!abc]", mode: :byte)
false
```
The `:codepoint` mode uses `String.codepoints/1` for tokenization, while
the `:byte` mode uses `:binary.bin_to_list/1`. If we tokenize our `"ā"`
subject we can see the two functions produce different amounts of tokens:
```
iex> String.codepoints("ā")
["ā"]
iex> :binary.bin_to_list("ā")
[196, 129]
```
If you are dealing with user input from forms this is likely not
something you will encounter and can keep the default value of
`:codepoint`.
* `:on_pattern_error` - What to do when the pattern is invalid. The
options are:
* `:fail` (default) - Simliar to case statements in Bash where an
invalid pattern won't match the subject, simply fail the match and return
`false`
* `:return` - Returns an `{:error, error}` tuple
* `:raise` - Raise an error
"""
@spec match?(binary(), binary()) :: boolean()
@spec match?(binary(), binary(), keyword()) :: boolean() | {:error, String.t()}
def match?(subject, pattern, opts \\ []) do
case Keyword.get(opts, :mode) do
:bash -> Bash.match?(subject, pattern, opts)
:byte -> Byte.match?(subject, pattern, opts)
_ -> Codepoint.match?(subject, pattern, opts)
end
end
@doc """
Checks if the given pattern is a valid unix-style wildcard pattern. The most
common invalid patterns arise because of invalid escape sequences.
It supports all of the usual wildcard pattern mechanisms:
- `*` matches none or many tokens
- `?` matches exactly one token
- `[abc]` matches a set of tokens
- `[a-z]` matches a range of tokens
- `[!...]` matches anything but a set of tokens
## Examples
iex> Wild.valid_pattern?("fo[a-z]b?r")
true
iex> Wild.valid_pattern?("\\a")
false
"""
@spec valid_pattern?(binary()) :: boolean()
def valid_pattern?(pattern) do
tokenize_result = Codepoint.tokenize_pattern(pattern)
Kernel.match?({:ok, _}, tokenize_result)
end
end