Chapter 5

Strings and characters

Strings are the basic communication medium for Unix processes, so a Unix programming environment must have reasonable facilities for manipulating them. Scsh provides a powerful set of procedures for processing strings and characters. Besides the the facilities described in this chapter, scsh also provides

• Regular expressions (chapter 6)
  A complete regular-expression system.

• Field parsing, delimited record I/O and the awk loop (chapter 8)
  These procedures let you read in chunks of text delimited by selected characters, and parse each record into fields based on regular expressions (for example, splitting a string at every occurrence of colon or white-space). The awk form allows you to loop over streams of these records in a convenient way.

• The SRFI-13 string libraries
  This pair of libraries contains procedures that create, fold, iterate over, search, compare, assemble, cut, hash, case-map, and otherwise manipulate strings. They are provided by the string-lib and string-lib-internals packages, and are also available in the default scsh package.

  More documentation on these procedures can be found at URLs

  http://srfi.schemers.org/srfi-13/srfi-13.html
  http://srfi.schemers.org/srfi-13/srfi-13.txt

• The SRFI-14 character-set library
  This library provides a set-of-characters abstraction, which is frequently useful when searching, parsing, filtering or otherwise operating on strings and character data. The SRFI is provided by the char-set-lib package; it's bindings are also available in the default scsh package.

  More documentation on this library can be found at URLs

  http://srfi.schemers.org/srfi-14/srfi-14.html
  http://srfi.schemers.org/srfi-14/srfi-14.txt

5.1  Manipulating file names

These procedures do not access the file-system at all; they merely operate on file-name strings. Much of this structure is patterned after the gnu emacs design. Perhaps a more sophisticated system would be better, something like the pathname abstractions of COMMON LISP or MIT Scheme. However, being Unix-specific, we can be a little less general.

5.1.1  Terminology

These procedures carefully adhere to the POSIX standard for file-name resolution, which occasionally entails some slightly odd things. This section will describe these rules, and give some basic terminology.

A file-name is either the file-system root (``/''), or a series of slash-terminated directory components, followed by a a file component. Root is the only file-name that may end in slash. Some examples:

Note that the relative filename src/des/main.c and the absolute filename /src/des/main.c are distinguished by the presence of the root component "" in the absolute path.

Multiple embedded slashes within a path have the same meaning as a single slash. More than two leading slashes at the beginning of a path have the same meaning as a single leading slash -- they indicate that the file-name is an absolute one, with the path leading from root. However, POSIX permits the OS to give special meaning to two leading slashes. For this reason, the routines in this section do not simplify two leading slashes to a single slash.

A file-name in directory form is either a file-name terminated by a slash, e.g., ``/src/des/'', or the empty string, ``''. The empty string corresponds to the current working directory, whose file-name is dot (``.''). Working backwards from the append-a-slash rule, we extend the syntax of POSIX file-names to define the empty string to be a file-name form of the root directory ``/''. (However, ``/'' is also acceptable as a file-name form for root.) So the empty string has two interpretations: as a file-name form, it is the file-system root; as a directory form, it is the current working directory. Slash is also an ambiguous form: / is both a directory-form and a file-name form.

The directory form of a file-name is very rarely used. Almost all of the procedures in scsh name directories by giving their file-name form (without the trailing slash), not their directory form. So, you say ``/usr/include'', and ``.'', not ``/usr/include/'' and ``''. The sole exceptions are file-name-as-directory and directory-as-file-name, whose jobs are to convert back-and-forth between these forms, and file-name-directory, whose job it is to split out the directory portion of a file-name. However, most procedures that expect a directory argument will coerce a file-name in directory form to file-name form if it does not have a trailing slash. Bear in mind that the ambiguous case, empty string, will be interpreted in file-name form, i.e., as root.

5.1.2  Procedures

These predicates return true if the string is in directory form, or file-name form (see the above discussion of these two forms). Note that they both return true on the ambiguous case of empty string, which is both a directory (current working directory), and a file name (the file-system root).

File name ...-directory? ...-non-directory? "src/des" #f #t "src/des/" #t #f "/" #t #f "." #f #t "" #t #t

Convert a file-name to directory form. Basically, add a trailing slash if needed:

(file-name-as-directory "src/des")	"src/des/"
(file-name-as-directory "src/des/")	"src/des/"
., /, and "" are special:
(file-name-as-directory ".")	""
(file-name-as-directory "/")	"/"
(file-name-as-directory "")	"/"

Convert a directory to a simple file-name. Basically, kill a trailing slash if one is present:

(directory-as-file-name "foo/bar/")	"foo/bar"
/ and "" are special:
(directory-as-file-name "/")	"/"
(directory-as-file-name "")	"." (i.e., the cwd)

Does fname begin with a root or ~ component? (Recognising ~ as a home-directory specification is an extension of POSIX rules.)

(file-name-absolute? "/usr/shivers")	#t
(file-name-absolute? "src/des")	#f
(file-name-absolute? "~/src/des")	#t
Non-obvious case:
(file-name-absolute? "")	#t (i.e., root)

Return the directory component of fname in directory form. If the file-name is already in directory form, return it as-is.

(file-name-directory "/usr/bdc")	"/usr/"
(file-name-directory "/usr/bdc/")	"/usr/bdc/"
(file-name-directory "bdc/.login")	"bdc/"
(file-name-directory "main.c")	""
Root has no directory component:
(file-name-directory "/")	""
(file-name-directory "")	""

Return non-directory component of fname.

(file-name-nondirectory "/usr/ian")	"ian"
(file-name-nondirectory "/usr/ian/")	""
(file-name-nondirectory "ian/.login")	".login"
(file-name-nondirectory "main.c")	"main.c"
(file-name-nondirectory "")	""
(file-name-nondirectory "/")	"/"

Split a file-name into its components.

(split-file-name "src/des/main.c")

("src" "des" "main.c")

(split-file-name "/src/des/main.c")

("" "src" "des" "main.c")

(split-file-name "main.c")

("main.c")

(split-file-name "/")

("")

Inverse of split-file-name.

(path-list->file-name '("src" "des" "main.c")) 
    ==>  "src/des/main.c"
(path-list->file-name '("" "src" "des" "main.c"))
    ==>  "/src/des/main.c"
[0]
Optional dir arg anchors relative path-lists:
(path-list->file-name '("src" "des" "main.c")
                      "/usr/shivers")
    ==>  "/usr/shivers/src/des/main.c"

The optional dir argument is usefully (cwd).

Return the file-name's extension.

(file-name-extension "main.c")	".c"
(file-name-extension "main.c.old")	".old"
(file-name-extension "/usr/shivers")	""

Weird cases:
(file-name-extension "foo.")	"."
(file-name-extension "foo..")	"."

Dot files are not extensions:
(file-name-extension "/usr/shivers/.login")	""

Return everything but the extension.

(file-name-sans-extension "main.c")	"main"
(file-name-sans-extension "main.c.old")	"main.c""
(file-name-sans-extension "/usr/shivers")
"/usr/shivers"

Weird cases:
(file-name-sans-extension "foo.")	"foo"
(file-name-sans-extension "foo..")	"foo."
Dot files are not extensions:
(file-name-sans-extension "/usr/shivers/.login")
"/usr/shivers/.login

Note that appending the results of file-name-extension and file-name-sans-extension in all cases produces the original file-name.

Let f be (file-name-nondirectory fname). This function returns the three values:

• (file-name-directory fname)

• (file-name-sans-extension f))

• (file-name-extension f)

The inverse of parse-file-name, in all cases, is string-append. The boundary case of / was chosen to preserve this inverse.

This procedure replaces fname's extension with ext. It is exactly equivalent to

(string-append (file-name-sans-extension fname) ext)

Removes leading and internal occurrences of dot. A trailing dot is left alone, as the parent could be a symlink. Removes internal and trailing double-slashes. A leading double-slash is left alone, in accordance with POSIX. However, triple and more leading slashes are reduced to a single slash, in accordance with POSIX. Double-dots (parent directory) are left alone, in case they come after symlinks or appear in a /../machine/... ``super-root'' form (which POSIX permits).

• Do ~ expansion.

• If dir is given, convert a relative file-name to an absolute file-name, relative to directory dir.

Resolve and simplify the file-name.

Convert file-name fname into an absolute file name, relative to directory dir, which defaults to the current working directory. The file name is simplified before being returned.

This procedure does not treat a leading tilde character specially.

home-dir returns user's home directory. User defaults to the current user.

(home-dir)	"/user1/lecturer/shivers"
(home-dir "ctkwan")	"/user0/research/ctkwan"

Returns file-name fname relative to user's home directory; user defaults to the current user.

(home-file "man")	"/usr/shivers/man"
(home-file "fcmlau" "man")	"/usr/fcmlau/man"

The general substitute-env-vars string procedure, defined in the previous section, is also frequently useful for expanding file-names.

5.2  Other string manipulation facilities

Replace occurrences of environment variables with their values. An environment variable is denoted by a dollar sign followed by alphanumeric chars and underscores, or is surrounded by braces.

(substitute-env-vars "$USER/.login")
"shivers/.login"
(substitute-env-vars "${USER}_log")	"shivers_log"

5.3  ASCII encoding

These are identical to char->integer and integer->char except that they use the ASCII encoding.

5.4  Character predicates

Each of these predicates tests for membership in one of the standard character sets provided by the SRFI-14 character-set library. Additionally, the following redundant bindings are provided for R5RS compatibility:

R5RS name	scsh definition
char-alphabetic?	char-letter+digit?
char-numeric?	char-digit?
char-alphanumeric?	char-letter+digit?

5.5  Deprecated character-set procedures

The SRFI-13 character-set library grew out of an earlier library developed for scsh. However, the SRFI standardisation process introduced incompatibilities with the original scsh bindings. The current version of scsh provides the library obsolete-char-set-lib, which contains the old bindings found in previous releases of scsh. The following table lists the members of this library, along with the equivalent SRFI-13 binding. This obsolete library is deprecated and not open by default in the standard scsh environment; new code should use the SRFI-13 bindings.

Old obsolete-char-set-lib	SRFI-13 char-set-lib
char-set-members	char-set->list
chars->char-set	list->char-set
ascii-range->char-set	ucs-range->char-set (not exact)
predicate->char-set	char-set-filter (not exact)
char-set-every?	char-set-every
char-set-any?	char-set-any
char-set-invert	char-set-complement
char-set-invert!	char-set-complement!
char-set:alphabetic	char-set:letter
char-set:numeric	char-set:digit
char-set:alphanumeric	char-set:letter+digit
char-set:control	char-set:iso-control

Note also that the ->char-set procedure no longer handles a predicate argument.