Icicles - Special Characters in Input Patterns

iciclesimage
Previous: Icicles - Inserting a Regexp from a Variable or RegisterIciclesIciclesIndexNext: Icicles - S-RET

Special Characters in Input Patterns

Regular-expression syntax treats some characters specially, but some of these special characters have another special meaning in Emacs when used with file-name inputs. What about the conflict between interpreting characters such as `$’, `\’, `.’, `?’, and `*’ as 1) regexp special characters and 2) special characters for file-name input? For example, when inputting a file name, should `*’ be treated as a regexp multiple-occurrences operator or as a file-name wildcard?

In Emacs file-name input:

Icicles handles the conflict by interpreting such characters as regexp special characters only during input completion and cycling – and then only if you do not escape them (with `\’). If present in the input when you finally accept it (using ‘RET’), they take on their normal Emacs meanings for file-name input: environment-variable prefix, wildcard, directory abbreviation, or directory separator.

That is, whenever there is a potential conflict of interpretation, the regexp meaning is used for completion and cycling, and the standard interpretation for file-name input is used for accepting the input. So, for example, to get the wildcard interpretation of `*’, just forego regexp completion and cycling. And vice versa: forego the wildcard interpretation to use regexp completion and cycling.

This is in any case the behavior of vanilla Emacs as well. If, in vanilla Emacs, you use `ici*’ or `ici*.el’ as input to ‘find-file’ and hit ‘TAB’, there is no completion available. File-name globbing and completion are independent.

Note: Because `?’ is useful in regexp syntax, the standard Emacs minibuffer binding of `?’, which just displays the completion-candidates list, is not used in Icicles. In Icicles, `?’ self-inserts in the minibuffer, like any other printable character. (Use ‘TAB’ or ‘S-TAB’ to display the list.) In standard Emacs, you must quote `?’ or copy-and-paste it, to insert it in the minibuffer for use as a file-name wildcard.

The interpretation conflict for `\’ (on MS Windows) is not really a problem, anyway. Although you cannot use a backslash (`\’) as a directory separator during apropos completion and cycling, you can always use a slash (`/’) instead – even on MS Windows. The best practice is to just break with MS-Windows syntax, and get in the habit of always using `/as the directory-separator character.

But what if you copy an absolute filename from some MS Windows application, so it has backslashes, and you want to use it in Emacs? You can go ahead and paste it in the minibuffer for file-name completion, as long as you are not doing regexp completion. You can (a) use prefix completion with it, (b) use `C-`’ to turn on regexp quoting for apropos completion (so you can complete a substring), or (c) change the backslashes to slashes.

Even if you always use only slash, not backslash, as a directory separator when inputting, however, it is possible that you could run into some trouble on MS Windows – you might (knowingly or not) use `\’ as a directory separator in the values of environment variables that you use as part of file-name input. If you are regexp completing, then those backslashes will be treated as regexp escapes. So you should use only non-regexp completion with input that includes environment variables whose expansions might include backslashes.

The interpretation conflict for `$’ is also not a real problem. You can get the effect of both interpretations of `$’ at the same time, because Icicles recognizes that `$’ at the end of input cannot be an environment-variable prefix. This means, for example, that you can use a pattern such as `$HOME.*t$’ to match the files in your home directory whose names end in ‘t’. The first `$’ here is not treated specially during regexp matching and cycling; the environment variable `$HOME’ is expanded by the shell to a directory name. The second `$’ is treated as the regexp special character that matches at the end of a line. When using environment variables, you can also enclose them in braces: `${HOME}’, for example.

Note: Starting with Emacs 23, if option `icicle-TAB-completion-methods’ includes ‘vanilla’, and you choose ‘vanilla’ completion for ‘TAB’ (by cycling using `C-(’ or by customizing ‘icicle-TAB-completion-methods’ to use ‘vanilla’ as the default), then Icicles ‘TAB’ completion will complete an environment variable during file-name completion. This is in addition to the traditional shell expansion of a variable when you hit ‘RET’.

Tip: Because slash (`/’) is about the only non-word syntax character that is likely to appear in file-name completions, you can usually use `\W$to match only directories (by matching the `/’ at the end of their names). `\W’ is the regexp pattern that matches any character that does not appear in words. For example, you can use `${HOME}\W$’ to match all direct subdirectories in your home directory.

Quoting (Escaping) Special Characters

You can toggle interpretation vs escaping of regexp special characters at any time, using `C-`’ in the minibuffer (command ‘icicle-toggle-regexp-quote’). Escaping special characters this way means they are no longer special; they simply match themselves. This has the effect of reducing apropos completion to simple substring completion. If you never want to use regexp matching (not recommended!), you can customize user option `icicle-regexp-quote-flag’, setting it to non-‘nil’.

Apropos (regexp) completion contains literal substring completion as a (common) special case. Sometimes you want to use regexp completion, but you also want to match a literal substring that contains special characters. You can of course quote (escape) each of these characters by hand. Alternatively, you can use `M-%’ (‘icicle-regexp-quote-input’) to quote the text that you want to match literally. If the region is active, then it is quoted; otherwise, your entire minibuffer input is quoted.

Note that if a substring that you want to match literally can occur anywhere in the target completion candidate, then it is simpler to just use progressive completion. Quoting a literal substring is useful when the overall regexp requires it to be at a certain location in the target.

See Also:


Previous: Icicles - Inserting a Regexp from a Variable or RegisterIciclesIciclesIndexNext: Icicles - S-RET

DrewsElispLibraries referenced here: Lisp:icicles.el

CategoryCommands CategoryCompletion CategoryRegexp CategoryDirectories CategoryFiles CategoryProgrammerUtils CategoryCode