man scan (3): Parse string using conversion specifiers in the style of sscanf

SYNOPSIS

scan string format ?varName varName ...?

INTRODUCTION

This command parses fields from an input string in the same fashion as the ANSI C sscanf procedure and returns a count of the number of conversions performed, or -1 if the end of the input string is reached before any conversions have been performed. String gives the input to be parsed and format indicates how to parse it, using % conversion specifiers as in sscanf. Each varName gives the name of a variable; when a field is scanned from string the result is converted back into a string and assigned to the corresponding variable. If no varName variables are specified, then scan works in an inline manner, returning the data that would otherwise be stored in the variables as a list. In the inline case, an empty string is returned when the end of the input string is reached before any conversions have been performed.

DETAILS ON SCANNING

Scan operates by scanning string and format together. If the next character in format is a blank or tab then it matches any number of white space characters in string (including zero). Otherwise, if it isn't a % character then it must match the next character of string. When a % is encountered in format, it indicates the start of a conversion specifier. A conversion specifier contains up to four fields after the %: a *, which indicates that the converted value is to be discarded instead of assigned to a variable; a XPG3 position specifier; a number indicating a maximum field width; a field size modifier; and a conversion character. All of these fields are optional except for the conversion character. The fields that are present must appear in the order given above.

When scan finds a conversion specifier in format, it first skips any white-space characters in string (unless the specifier is [ or c). Then it converts the next input characters according to the conversion specifier and stores the result in the variable given by the next argument to scan.

If the % is followed by a decimal number and a $, as in ``%2$d'', then the variable to use is not taken from the next sequential argument. Instead, it is taken from the argument indicated by the number, where 1 corresponds to the first varName. If there are any positional specifiers in format then all of the specifiers must be positional. Every varName on the argument list must correspond to exactly one conversion specifier or an error is generated, or in the inline case, any position can be specified at most once and the empty positions will be filled in with empty strings.

The following conversion characters are supported:

d: The input field must be a decimal integer. It is read in and the value is stored in the variable as a decimal string. If the l or L field size modifier is given, the scanned value will have an internal representation that is at least 64-bits in size.
o: The input field must be an octal integer. It is read in and the value is stored in the variable as a decimal string. If the l or L field size modifier is given, the scanned value will have an internal representation that is at least 64-bits in size. If the value exceeds MAX_INT (017777777777 on platforms using 32-bit integers when the l and L modifiers are not given), it will be truncated to a signed integer. Hence, 037777777777 will appear as -1 on a 32-bit machine by default.
x: The input field must be a hexadecimal integer. It is read in and the value is stored in the variable as a decimal string. If the l or L field size modifier is given, the scanned value will have an internal representation that is at least 64-bits in size. If the value exceeds MAX_INT (0x7FFFFFFF on platforms using 32-bit integers when the l and L modifiers are not given), it will be truncated to a signed integer. Hence, 0xFFFFFFFF will appear as -1 on a 32-bit machine.
u: The input field must be a decimal integer. The value is stored in the variable as an unsigned decimal integer string. If the l or L field size modifier is given, the scanned value will have an internal representation that is at least 64-bits in size.
i: The input field must be an integer. The base (i.e. decimal, octal, or hexadecimal) is determined in the same fashion as described in expr. The value is stored in the variable as a decimal string. If the l or L field size modifier is given, the scanned value will have an internal representation that is at least 64-bits in size.
c: A single character is read in and its binary value is stored in the variable as a decimal string. Initial white space is not skipped in this case, so the input field may be a white-space character. This conversion is different from the ANSI standard in that the input field always consists of a single character and no field width may be specified.
s: The input field consists of all the characters up to the next white-space character; the characters are copied to the variable.
e or f or g: The input field must be a floating-point number consisting of an optional sign, a string of decimal digits possibly containing a decimal point, and an optional exponent consisting of an e or E followed by an optional sign and a string of decimal digits. It is read in and stored in the variable as a floating-point string.
[chars]: The input field consists of any number of characters in chars. The matching string is stored in the variable. If the first character between the brackets is a ] then it is treated as part of chars rather than the closing bracket for the set. If chars contains a sequence of the form a-b then any character between a and b (inclusive) will match. If the first or last character between the brackets is a -, then it is treated as part of chars rather than indicating a range.
[^chars]: The input field consists of any number of characters not in chars. The matching string is stored in the variable. If the character immediately following the ^ is a ] then it is treated as part of the set rather than the closing bracket for the set. If chars contains a sequence of the form a-b then any character between a and b (inclusive) will be excluded from the set. If the first or last character between the brackets is a -, then it is treated as part of chars rather than indicating a range.
n: No input is consumed from the input string. Instead, the total number of characters scanned from the input string so far is stored in the variable.

The number of characters read from the input for a conversion is the largest number that makes sense for that particular conversion (e.g. as many decimal digits as possible for %d, as many octal digits as possible for %o, and so on). The input field for a given conversion terminates either when a white-space character is encountered or when the maximum field width has been reached, whichever comes first. If a * is present in the conversion specifier then no variable is assigned and the next scan argument is not consumed.

DIFFERENCES FROM ANSI SSCANF

The behavior of the scan command is the same as the behavior of the ANSI C sscanf procedure except for the following differences:

[1]: %p conversion specifier is not currently supported.
[2]: For %c conversions a single character value is converted to a decimal string, which is then assigned to the corresponding varName; no field width may be specified for this conversion.
[3]: The h modifier is always ignored and the l and L modifiers are ignored when converting real values (i.e. type double is used for the internal representation).
[4]: If the end of the input string is reached before any conversions have been performed and no variables are given, an empty string is returned.

EXAMPLES

Parse a simple color specification of the form #RRGGBB using hexadecimal conversions with field sizes:

set string "#08D03F"
scan $string "#%2x%2x%2x" r g b

Parse a HH:MM time string, noting that this avoids problems with octal numbers by forcing interpretation as decimals (if we did not care, we would use the %i conversion instead):

set string "08:08"   ;# *Not* octal!
if {[scan $string "%d:%d" hours minutes] != 2} {
   error "not a valid time string"
}
# We have to understand numeric ranges ourselves...
if {$minutes < 0 || $minutes > 59} {
   error "invalid number of minutes"
}

Break a string up into sequences of non-whitespace characters (note the use of the %n conversion so that we get skipping over leading whitespace correct):

set string " a string {with braced words} + leading space "
set words {}
while {[scan $string %s%n word length] == 2} {
   lappend words $word
   set string [string range $string $length end]
}

Parse a simple coordinate string, checking that it is complete by looking for the terminating character explicitly:

set string "(5.2,-4e-2)"
# Note that the spaces before the literal parts of
# the scan pattern are significant, and that ")" is
# the Unicode character \u0029
if {
   [scan $string " (%f ,%f %c" x y last] != 3
   || $last != 0x0029
} then {
   error "invalid coordinate string"
}
puts "X=$x, Y=$y"

KEYWORDS

conversion specifier, parse, scan