Standard Library

Ruby comes “out of the box” with a large and useful library of modules and classes. This chapter contains a sampling of the more useful of these.

Interestingly, and unlike some of the code in later chapters, all of these libraries are written in Ruby. You'll find the source in the lib subdirectory of the standard Ruby distribution.

class Complex < Numeric

require "complex"

v1 = Complex(2,3) → Complex(2, 3)
v2 = 2.im → Complex(0, 2)
v1 + v2 → Complex(2, 5)
v1 * v2 → Complex(-6, 4)
v2**2 → Complex(-4, 0)
Math.sin(v1) → Complex(9.154499147, -4.16890696)
v1 < v2 → false
v2**2 == -4 → true

Complex constants

`Complex::I`	0 +1_i

Complex class methods

new

Complex.new( a, b ) → aComplex

Returns a + b_i.

In addition to the Complex.new constructor, the Complex library defines the method Numeric.im, such that aNumeric.im returns 0 + aNumeric_i. Complex numbers are also constructed using the global method Complex, which takes one or two arguments. The value it returns depends on the type of its arguments:

a	b	Result
Number	Number	a + b_i
Complex	0	a
Complex	Complex	`Complex( a.real - b.image, a.image + b.real )`
Number	Complex	`Complex( a - b.image, b.real )`

Complex instance methods

Arithmetic operations

Performs various arithmetic operations on ref.

ref	`+`	aNumeric → aComplex	Addition
ref	`-`	aNumeric → aComplex	Subtraction
ref	`*`	aNumeric → aComplex	Multiplication
ref	`/`	aNumeric → aComplex	Division
ref	`%`	aNumeric → aComplex	Remainder
ref	`**`	aNumeric → aComplex	Exponentiation (real and complex power)

<=>

ref <=> other → -1, 0, +1

Returns ref.abs <=> other.abs.

==

ref == anObject → true or false

If anObject is a complex number, returns true if its real and imaginary parts match ref. If anObject is a simple number, returns true if ref.real equals anObject and ref.image is zero. Otherwise, attempts to coerce anObject to a complex number and compares the result.

abs

ref.abs → aFloat

Absolute value.

abs2

ref.abs2 → aFloat

Square of absolute value.

arg

ref.arg → aFloat

Argument (angle from (1,0)).

conjugate

ref.conjugate → aComplex

Complex conjugate.

image

ref.image → aNumeric

The imaginary part of ref.

polar

ref.polar → anArray

Returns the two-element array: [c.abs, c.arg].

real

ref.real → aNumeric

The real part of ref.

to_f

ref.to_f → aComplex

Returns Complex(real.to_f, image.to_f).

to_i

ref.to_i → aComplex

Returns Complex(real.to_i, image.to_i).

to_r

ref.to_r → aComplex

Returns Complex(real.to_r, image.to_r), converting both parts of the complex to a rational number.

to_s

ref.to_s → aString

String representation of ref.

In addition, the Math functions sqrt, exp, cos, sin, tan, log, log10, and atan2 are extended to support a Complex argument.

class Date < Object

require 'date'

d = Date.new(2000, 3, 31) → #<Date: 2451635,2299161>
[d.year, d.yday, d.wday] → [2000, 91, 5]
[d.month, d.mday] → [3, 31]
[d.cwyear, d.cweek, d.cwday] → [2000, 13, 5]
[d.jd, d.mjd] → [2451635, 51634.5]
(d << 1).to_s → "2000-02-29"
d.succ.to_s → "2000-04-01"
(d + 100).to_s → "2000-07-09"
d.leap? → true
Date.new(2000, 3, -10).to_s → "2000-03-22"
d1 = Date.neww(2000, 13, 7) → #<Date: 2451637,2299161>
d1.to_s → "2000-04-02"
[d1.cwday, d1.wday] → [7, 0]

The date library implements class Date, which provides a comprehensive set of facilities for storing, manipulating, and converting dates. To document its options, we need to take a brief historical detour to establish some vocabulary.

Internally a date is stored as a Julian day number, the number of days since midday, January 1st, 4713 BCE. (In the code, you may find references to the year -4712. As astronomical dates include a year zero, 4713 BCE is the same year as -4712.) The rules for converting a Julian day number to a calendar date are complicated because the Romans estimated the length of a year incorrectly. In the Julian calendar (often called Old Style, or O.S.), every year divisible by 4 is a leap year. The Date class has options to convert dates using this as an assumption.

By the sixteenth century, the inaccuracies in this measurement had become apparent. An edict from Pope Gregory XIII in 1582 created the New Style (N.S.) or Gregorian calendar, where years divisible by 100 were no longer leap years unless they were also divisible by 400. This system was adopted by most Catholic countries immediately, but religious differences held up a wider adoption. England (and several other countries) switched in 1752, with some countries following later. The Date class allows you to determine whether to implement the cutover in 1582 (the Date::ITALY option), 1752 (Date::ENGLAND), or another date of your choosing.

The Date class also provides conversions to Modified Julian Day (MJD) numbers. MJD values count from midnight, November 17, 1858. Because these values count from midnight, not midday, there is a half-day added to the conversion factor.

The descriptions that follow use the abbreviations listed in Table 24.1.

Class Date exports the constant arrays Date::MONTHNAMES and Date::DAYNAMES, which can be indexed by mon and wday values to return the corresponding English names.

The Date class also provides low-level date-conversion methods:

civil_to_jd
commercial_to_jd
ordinal_to_jd
mjd_to_jd
jd_to_mjd
jd_to_civil
jd_to_commercial
jd_to_ordinal

These methods perform limited error checking of their parameters, and are not documented here. The somewhat confusingly named exist..? routines perform conversions from different formats into a Julian day number with error checking. These routines also automatically normalize their parameters.

Table 24.1 : Abbreviations used describing dates
Field	Meaning
cwday	An ISO 8601 calendar weekday. 1 is Monday, 7 is Sunday.
cweek	An ISO 8601 calendar week. Week 1 is the week containing the first Thursday (or equivalently the week that contains January 4th).
cwyear	An ISO 8601 calendar-week-based year. May be different from year, as it rolls forward only on a Monday.
jd	The Julian day number—the number of days since January 1st, 4713 BCE.
mday	The day of the month (1..31).
mjd	A modified Julian day number.
mon	The month of the year (1..12).
sg	The start of the Gregorian correction: `Date::ITALY` (the default) for 1582, `Date::ENGLAND` for 1752, or `JULIAN`, meaning no correction. You may also provide an arbitrary Julian day number for this parameter, in which case the correction will start from this date.
wday	The day of the week (0 is Sunday).
week	The week number into a year (1..53).
yday	The day into the year (1..366).
year	A year (1966, 2001, and the like).

mixins

Comparable: <, <=, ==, >=, >, between?

Date class methods

exist2?: Date.exist2?( year, yday, sg=Date::ITALY) → jd; Converts a year and yday into a Julian day number, returning nil on error.
exist?: Date.exist?( year, mon, mday, sg=Date::ITALY) → jd; Converts a year, mon, and mday into a Julian day number, or nil if the parameters are invalid.
existw?: Date.existw?( cyear, cweek, cwday, sg=Date::ITALY) → jd; Converts a cyear, cweek, and cwday into a Julian day number.
gregorian_leap?: Date.gregorian_leap?( year ) → true or false; If year does not end with “00”, returns true if year is divisible by 4, otherwise returns true if year is divisible by 400.
julian_leap?: Date.julian_leap?( year ) → true or false; Returns true if year is divisible by 4.
leap?: Date.leap?( year ) → true or false; Synonym for Date.gregorian_leap?.
new: Date.new( year=-4712, mon=1, mday=1, sg=Date::ITALY) → aNewDate; Returns a Date for the given year, mon, and mday. If mon is negative, it counts back from the end of the year. If mday is negative, it counts back from the end of the month.
new1: Date.new1( jd, sg=Date::ITALY) → aNewDate; Creates a Date corresponding to the given Julian day number.
new2: Date.new2( year=-4712, yday=1, sg=Date::ITALY) → aNewDate; Returns a Date for the given year and yday. If yday is negative, it counts back from the end of the year.
new3: Date.new3( year=-4712, mon=1, mday=1, sg=Date::ITALY) → aNewDate; Synonym for Date.new.
neww: Date.neww( cyear=1582, cweek=41, cwday=5, sg=Date::ITALY) → aNewDate; Returns a Date for the given cyear, cweek, and cwday. If cweek is negative, it counts back from the end of the year. If cwday is negative, it counts back from the end of the week.
today: Date.today( sg=Date::ITALY) → aNewDate; Returns a Date for today.

Date instance methods

Accessors: ref.year → year; ref.yday → yday; ref.mjd → mjd; ref.mon → mon; ref.month → mon; ref.mday → mday; ref.day → mday; ref.cwyear → cwyear; ref.cweek → cweek; ref.cwday → cwday; ref.wday → wday; Returns the given component of ref as a number.
+: ref + anInteger → aNewDate; Returns a new Date anInteger days from ref.
-: ref - anInteger → aNewDate; ref - anOtherDate → anInteger; The first form returns a new Date anInteger days before ref. The second form returns the number of days between ref and anOtherDate.
<<: ref << anInteger → aNewDate; Returns a new Date formed by subtracting anInteger months to ref, adjusting the mday value back to the last day of the month if it otherwise exceeds it.
<=>: ref <=> anOther → -1, 0, +1; anOther must be a Numeric, in which case it is treated as a Julian day number, or a Date. Returns -1, 0, +1 if ref is less than, equal to, or greater than anOther. See module Comparable.
===: ref === anOther → true or false; anOther must be a Numeric, in which case it is treated as a Julian day number, or a Date. Returns true if the Julian day number of anOther is the same as ref.
>>: ref >> anInteger → aNewDate; Returns a new Date formed by adding anInteger months to ref, adjusting the mday value back to the last day of the month if it otherwise exceeds it.
downto: ref.downto( aDateMin ) {| date | block } → ref; Invokes block with dates from ref down to aDateMin.
england: ref.england → aDate; Equivalent to ref.newsg(Date::ENGLAND).
gregorian: ref.gregorian → aDate; Equivalent to ref.newsg(Date::GREGORIAN).
italy: ref.italy → aDate; Equivalent to ref.newsg(Date::ITALY).
jd: ref.jd → jd; Returns the Julian day number for ref.
julian: ref.julian → aDate; Equivalent to ref.newsg(Date::JULIAN).
leap?: ref.leap? → true or false; Returns true if ref falls within a leap year.
mjd: ref.mjd → mjd; Returns the Julian day number of ref converted to a modified Julian day number.
newsg: ref.newsg( sg=Date::ITALY ) → aNewDate; Returns a new Date.
next: ref.next → aNewDate; Synonym for ref.succ.
ns?: ref.ns? → true or false; Returns true if ref falls in the period of New Style dates.
os?: ref.os? → true or false; Returns true if ref falls in the period of Old Style dates.
sg: ref.sg → anInteger; Returns the Julian day number of the start of New Style dates for ref.
step: ref.step( aDateLimit, step ) {| date | block } → ref; Invokes block with dates starting at ref, incrementing by step days, ending at the first date greater than aDateLimit (less than for a negative step).
succ: ref.succ → aNewDate; Returns the date of ref plus one day.
to_s: ref.to_s → aString; Returns self as “year-mon-mday.”
upto: ref.upto( aDateMax ) {| date | block } → ref; Invokes block with dates from ref to aDateMax.

library English

require "English"

$OUTPUT_FIELD_SEPARATOR = ' -- '
"waterbuffalo" =~ /buff/
print $LOADED_FEATURES, $POSTMATCH, $PID, "\n"
print $", $', $$, "\n"

produces:

English.rb -- alo -- 32130 --
English.rb -- alo -- 32130 --

Include the English library file in a Ruby script, and you can reference the global variables such as $_ using less cryptic names, listed in the following table.

`$*`	$ARGV	`$"`	$LOADED_FEATURES
`$?`	$CHILD_STATUS	`$&`	$MATCH
`$<`	$DEFAULT_INPUT	`$.`	$NR
`$>`	$DEFAULT_OUTPUT	`$,`	$OFS
`$!`	$ERROR_INFO	`$\`	$ORS
`$@`	$ERROR_POSITION	`$\`	$OUTPUT_RECORD_SEPARATOR
`$;`	$FIELD_SEPARATOR	`$,`	$OUTPUT_FIELD_SEPARATOR
`$;`	$FS	`$$`	$PID
`$=`	$IGNORECASE	`$'`	$POSTMATCH
`$.`	$INPUT_LINE_NUMBER	$`	$PREMATCH
`$/`	$INPUT_RECORD_SEPARATOR	`$$`	$PROCESS_ID
`$~`	$LAST_MATCH_INFO	`$0`	$PROGRAM_NAME
`$+`	$LAST_PAREN_MATCH	`$/`	$RS
`$_`	$LAST_READ_LINE

module Find

require "find"
Find.find("/etc/passwd", "/var/spool/lp1", ".") do |f|
  Find.prune if f == "."
  puts f
end

produces:

/etc/passwd
/var/spool/lp1
/var/spool/lp1/status
/var/spool/lp1/lock
/var/spool/lp1/.seq

The Find module supports the top-down traversal of a set of file paths.

Find class methods

find: ref.find( [aName]* ) {| aFileName | block }; Calls the associated block with the name of every file and directory listed as arguments, then recursively on their subdirectories, and so on.
prune: ref.prune; Skips the current file or directory, restarting the loop with the next entry. If the current file is a directory, that directory will not be recursively entered. Meaningful only within the block associated with Find.find.

class File < IO

require 'ftools'

File.copy 'testfile', 'testfile1'     → true
File.compare 'testfile', 'testfile1'  → true

The FTools library adds several methods to the built-in File class. These methods are particularly useful to programs that move and copy files, such as installers.

File class methods

cmp: ref.cmp( name1, name2, verbose=false ) → true or false; Synonym for File.compare.
compare: ref.compare( name1, name2, verbose=false ) → true or false; Returns true only if the contents of files name1 and name2 are identical.
copy: ref.copy( fromName, toName, verbose=false ) → true or false; Equivalent to calling File.syscopy, but logs the attempt to $stderr if verbose is not false.
cp: ref.cp( fromName, toName, verbose=false ) → true or false; Synonym for File.copy.
install: ref.install( fromName, toName, aMode=nil, verbose=false ); Copies file fromName to file toName using File.syscopy, unless toName already exists and has the same content as fromName. Sets the mode of the resulting file to aMode unless aMode is nil.
makedirs: ref.makedirs( [dirName]* [, aBoolean] ); Creates the given directories, logging each attempt to $stderr if the last parameter is true. Creates any missing parent directories as required.
mkpath: ref.mkpath( [dirName]* [, aBoolean] ); Synonym for File.makedirs.
move: ref.move( fromName, toName, verbose=false ) → true or false; Effectively renames fromName to toName, logging to $stderr if verbose is not false.
mv: ref.mv( fromName, toName, verbose=false ) → true or false; Synonym for File.move.
rm_f: ref.rm_f( [fileName]* [, aBoolean] ) → anInteger; Synonym for File.safe_unlink (the name refers to the Unix rm -f command).
safe_unlink: ref.safe_unlink( [fileName]* [, aBoolean] ) → anInteger or nil; Unlinks (deletes) the given files, logging to $stderr if the last parameter is true. The method attempts to make all files writable before unlinking them, so no errors will occur deleting read-only files. Returns the number of files deleted, or nil on error.
syscopy: ref.syscopy( fromName, toName ) → true or false; Efficiently copies the file named fromName to toName. If toName names a directory, the destination will be a file in that directory with the same basename as fromName. After the copy, the file mode of toName will be the same as that of fromName. Returns true on success.

class GetoptLong < Object

# Call using "ruby example.rb --size 10k -v -q a.txt b.doc"

require 'getoptlong'

# specify the options we accept and initialize
# the option parser

opts = GetoptLong.new(
  [ "--size",    "-s",            GetoptLong::REQUIRED_ARGUMENT ],
  [ "--verbose", "-v",            GetoptLong::NO_ARGUMENT ],
  [ "--query",   "-q",            GetoptLong::NO_ARGUMENT ],
  [ "--check",   "--valid", "-c", GetoptLong::NO_ARGUMENT ]
)

# process the parsed options

opts.each do |opt, arg|
  puts "Option: #{opt}, arg #{arg.inspect}"
end

puts "Remaining args: #{ARGV.join(', ')}"

produces:

Option: --size, arg "10k"
Option: --verbose, arg ""
Option: --query, arg ""
Remaining args: a.txt, b.doc

Class GetoptLong supports GNU-style command-line option parsing. Options may be a minus sign (`-') followed by a single character, or two minus signs (`--') followed by a name (a long option). Long options may be abbreviated to their shortest unambiguous lengths.

A single internal option may have multiple external representations. For example, the option to control verbose output could be any of -v, --verbose, or --details. Some options may also take an associated value.

Each internal option is passed to GetoptLong as an array, containing strings representing the option's external forms and a flag. The flag (NO_ARGUMENT, REQUIRED_ARGUMENT, or OPTIONAL_ARGUMENT) specifies how GetoptLong is to associate an argument with the option.

If the environment variable POSIXLY_CORRECT is set, all options must precede nonoptions on the command line. Otherwise, the default behavior of GetoptLong is to reorganize the command line to put the options at the front. This behavior may be changed by setting GetoptLong#ordering= to one of the constants PERMUTE, REQUIRE_ORDER, or RETURN_IN_ORDER. POSIXLY_CORRECT may not be overridden.

GetoptLong constants

Per-option constants
`NO_ARGUMENT`	Flags an option that takes no argument.
`OPTIONAL_ARGUMENT`	A nonoption following this option will be used as this option's argument.
`REQUIRED_ARGUMENT`	This option must be followed by an argument.

Overall constants
`PERMUTE`	Options and their arguments will be shuffled to the front of the command line.
`REQUIRE_ORDER`	Options and their arguments must appear at the start of the command line. The first nonoption terminates option processing.
`RETURN_IN_ORDER`	Return options in the order in which they occur on the command line.

GetoptLong class methods

new: GetoptLong.new( [options]* ) → ref; Returns a new option parser. Any options are passed to ref.set_options.

GetoptLong instance methods

each

ref.each {| anOption, anArgument | block }

Loops calling GetoptLong#get, passing the returned option and argument to the associated block. The loop ends when get returns nil for anOption.

error?

ref.error? → anException

Returns an Exception object documenting any error that has occurred, or nil if there has not been an error.

error_message

ref.error_message → aString

Returns the text of the last error message.

get

ref.get → [ anOption, anArgument ]

Returns the next option, along with any associated argument. If there is no argument, nil is returned for anArgument. If there are no remaining unprocessed options, or if there is an error in option processing and quiet has been set, nil is returned for anOption. Otherwise, if there is an error, a message is written to $stderr and an exception (a subclass of StandardError) is raised.

The option string returned is the first option that was given in the corresponding array passed to set_options.

get_option

ref.get_option → [ anOption, anArgument ]

Synonym for GetoptLong#get.

ordering

ref.ordering → aFixnum

Returns the current ordering.

ordering=

ref.ordering = aFixnum

Sets the ordering to one of PERMUTE, REQUIRE_ORDER, or RETURN_IN_ORDER. Quietly ignored if the environment variable POSIXLY_CORRECT is set. Ordering may not be changed once option processing has been started.

quiet

ref.quiet → true or false

Returns the current value of the quiet attribute.

quiet=

ref.quiet = true or false

Sets the current value of the quiet attribute. If false, any errors encountered are reported to $stderr.

quiet?

ref.quiet? → true or false

Synonym for GetoptLong#quiet.

set_options

ref.set_options( [anOptArray]* ) → ref

Each parameter is an array specifying a single internal option. The array contains one or more strings specifying the external form(s) of the option, and one of the flags NO_ARGUMENT, OPTIONAL_ARGUMENT, or REQUIRED_ARGUMENT. See the sample code on at the start of the documentation for this class.

terminate

ref.terminate → ref

Terminates option processing. Any remaining arguments are written back to ARGV. This may be called from within a GetoptLong#each or on its own. For example, calling the following program using “ruby example.rb --size 10k -v -term -q a.txt b.doc” will leave the -q and filenames in ARGV.

require 'getoptlong'

opts = GetoptLong.new(
  [ "--size",    "-s",            GetoptLong::REQUIRED_ARGUMENT ],
  [ "--verbose", "-v",            GetoptLong::NO_ARGUMENT ],
  [ "--term",    "-t",            GetoptLong::NO_ARGUMENT ],
  [ "--query",   "-q",            GetoptLong::NO_ARGUMENT ],
  [ "--check",   "--valid", "-c", GetoptLong::NO_ARGUMENT ]
  )

opts.each do |opt, arg|
  puts "Option: #{opt}, arg #{arg.inspect}"
  opts.terminate if (opt == '--term')
end

puts "Remaining args: #{ARGV.join(', ')}"

produces:

Option: --size, arg "10k"
Option: --verbose, arg ""
Option: --term, arg ""
Remaining args: -q, a.txt, b.doc

terminated?

ref.terminated? → true or false

Returns true if option processing has been terminated.

module mkmf

The mkmf library is used by Ruby extension modules to help create Makefiles. When writing an extension, you create a program named “extconf.rb”, which may be as simple as:

require 'mkmf'
create_makefile("Test")

When run, this script will produce a Makefile suited to the target platform. mkmf contains several methods you can use to find libraries and include files and to set compiler flags.

For more information on creating extension modules, see Chapter 17, “Extending Ruby.”

mkmf constants

`PLATFORM`	varies	A constant string that describes the platform on which Ruby is running, such as “mswin32” or “i686-linux.”
`$CFLAGS`		Global variable for compiler flags.
`$LDFLAGS`		Global variable for linker flags.

mkmf instance methods

create_makefile

create_makefile( target )

Creates a Makefile for an extension named target. If this method is not called, no Makefile is created.

dir_config

dir_config( name )

Looks for directory configuration options for name given as arguments to this program or to the original build of Ruby. These arguments may be one of:

--with-name-dir=directory

--with-name-include=directory

--with-name-lib=directory

The given directories will be added to the appropriate search paths (include or link) in the Makefile.

find_library

find_library( name, function, [path]+ ) → true or false

Same as have_library, but will also search in the given directory paths.

have_func

have_func( function ) → true or false

If the named function exists in the standard compile environment, adds the directive -DHAVE_FUNCTION to the compile command in the Makefile and returns true.

have_header

have_header( header ) → true or false

If the given header file can be found in the standard search path, adds the directive -DHAVE_HEADER to the compile command in the Makefile and returns true.

have_library

have_library( library, function ) → true or false

If the given function exists in the named library, which must exist in the standard search path or in a directory added with dir_config, adds the library to the link command in the Makefile and returns true.

module ParseDate

The ParseDate module defines a single method, ParseDate.parsedate, which converts a date and/or time string into its constituents. It uses heuristics that handle a wide variety of date and time formats, including a subset of ISO 8601, Unix ctime, and most common written variants. The following table shows some examples.

String	Guess?	yy	mm	dd	hh	min	sec	zone	wd
1999-09-05 23:55:21+0900	F	1999	9	5	23	55	21	+0900	--
1983-12-25	F	1983	12	25	--	--	--	--	--
1965-11-10 T13:45	F	1965	11	10	13	45	--	--	--
10/9/75 1:30pm	F	75	10	9	13	30	--	--	--
10/9/75 1:30pm	T	1975	10	9	13	30	--	--	--
Mon Feb 28 17:15:49 CST 2000	F	2000	2	28	17	15	49	CST	1
Tue, 02-Mar-99 11:20:32 GMT	F	99	3	2	11	20	32	GMT	2
Tue, 02-Mar-99 11:20:32 GMT	T	1999	3	2	11	20	32	GMT	2
12-January-1990, 04:00 WET	F	1990	1	12	4	0	--	WET	--
4/3/99	F	99	4	3	--	--	--	--	--
4/3/99	T	1999	4	3	--	--	--	--	--
10th February, 1976	F	1976	2	10	--	--	--	--	--
March 1st, 84	T	1984	3	1	--	--	--	--	--
Friday	F	--	--	--	--	--	--	--	5

ParseDate class methods

parsedate: ParseDate.parsedate( aString, guessYear=false ) → [ year, mon, mday, hour, min, sec, zone, wday ]; Parses a string containing a date and/or a time, returning an array of Fixnum objects containing the various components. nil is returned for fields that cannot be parsed from aString. If the result contains a year that is less than 100 and guessYear is true, parsedate will return a year value equal to year plus 2000 if year is less than 69, year plus 1900 otherwise.

library profile

The profile library prints to $stderr a summary of the number of calls to, and the time spent in, each method in a Ruby program. The output is sorted by the total time spent in each method. Profiling can be enabled from the command line using the -rprofile option, or from within a source program by requiring the profile module.

require 'profile'
def ackerman(m, n)
  if m == 0 then  n+1
  elsif n == 0 and m > 0 then ackerman(m-1, 1)
  else ackerman(m-1, ackerman(m, n-1))
  end
end
ackerman(3,3)

produces:

time   seconds   seconds    calls  ms/call  ms/call  name
 74.17     2.47      2.47     2432     1.02    41.95  Object#ackerman
 17.42     3.05      0.58     3676     0.16     0.16  Fixnum#==
  5.71     3.24      0.19     2431     0.08     0.08  Fixnum#-
  2.70     3.33      0.09     1188     0.08     0.08  Fixnum#+
  0.00     3.33      0.00        1     0.00     0.00  Module#method_added
  0.00     3.33      0.00       57     0.00     0.00  Fixnum#>
  0.00     3.33      0.00        1     0.00  3330.00  #toplevel

class PStore < Object

The PStore class provides transactional, file-based persistent storage of Ruby objects. The following example stores two hierarchies in a PStore. The first, identified by the key “names”, is an array of Strings. The second, identified by “tree”, is a simple binary tree.

require "pstore"

class T
  def initialize(val, left=nil, right=nil)
    @val, @left, @right = val, left, right
  end
  def to_a
    [ @val, @left.to_a, @right.to_a ]
  end
end

store = PStore.new("/tmp/store")
store.transaction do
   store['names'] = [ 'Douglas', 'Barenberg', 'Meyer' ]
   store['tree']  = T.new('top',
                      T.new('A', T.new('B')),
                      T.new('C', T.new('D', nil, T.new('E'))))
end

# now read it back in

store.transaction do
   puts "Roots: #{store.roots.join(', ')}"
   puts store['names'].join(', ')
   puts store['tree'].to_a.inspect
end

produces:

Roots: names, tree
Douglas, Barenberg, Meyer
["top", ["A", ["B", [], []], []], ["C", ["D", [], ["E", [], []]], []]]

Each PStore can store several object hierarchies. Each hierarchy has a root, identified by a key (often a string). At the start of a PStore transaction, these hierarchies are read from a disk file and made available to the Ruby program. At the end of the transaction, the hierarchies are written back to the file. Any changes made to objects in these hierarchies are therefore saved on disk, to be read at the start of the next transaction that uses that file.

In normal use, a PStore object is created and then is used one or more times to control a transaction. Within the body of the transaction, any object hierarchies that had previously been saved are made available, and any changes to object hierarchies, and any new hierarchies, are written back to the file at the end.

PStore class methods

new: PStore.new( aFilename ) → aPStore; Returns a new PStore object associated with the given file. If the file exists, its contents must have been previously written by PStore.

PStore instance methods

[ ]: ref[ anObject ] → anOtherObject; Root Access—Returns the root of an object hierarchy identified by anObject. An exception is raised if anObject does not identify a root.
[ ]=: ref[ anObject ] = anOtherObject → anOtherObject; Root Creation—Sets anOtherObject as the base of the object hierarchy to be identified using anObject.
abort: ref.abort; Terminates this transaction, losing any changes made to the object hierarchies.
commit: ref.commit; Terminates the current transaction, saving the object hierarchies into the store's file.
path: ref.path → aString; Returns the name of the file associated with this store.
root?: ref.root?( anObject ) → true or false; Returns true if anObject is the key of a root in this store.
roots: ref.roots → anArray; Returns an array containing the keys of the root objects available in this store.
transaction: ref.transaction {| ref | block } → anObject; If the file associated with ref exists, reads in the object hierarchies from it. It then executes the associated block, passing in ref. The block may use this parameter to access the roots of the hierarchies and hence access the persistent objects. If the block calls PStore#abort, or if it raises an exception, no data is saved back to the associated file. Otherwise, if it invokes PStore#commit, or if it terminates normally, the object hierarchies are written back to the file. The value returned is the value returned by the block.

class Tempfile < IO

require "tempfile"

tf = Tempfile.new("afile")
tf.path → "/tmp/afile32146.0"
tf.puts("Cosi Fan Tutte") → nil
tf.close → nil
tf.open → #<File:0x40196fc8>
tf.gets → "Cosi Fan Tutte\n"
tf.close(true) → #<File:0x40196fc8>

Class Tempfile creates managed temporary files. Although they behave the same as any other IO objects, temporary files are automatically deleted when the Ruby program terminates. Once a Tempfile object has been created, the underlying file may be opened and closed a number of times in succession.

Tempfile does not directly inherit from IO. Instead, it delegates calls to a File object. From the programmer's perspective, apart from the unusual new, open, and close semantics, a Tempfile object behaves as if it were an IO object.

Tempfile class methods

new

Tempfile.new( basename, tmpdir=<see below> ) → ref

Constructs a temporary file in the given directory. The file name is built by concatenating basename, the current process id and (as an extension) a unique sequence number. If the tmpdir parameter is not supplied, it defaults to the value of one of the environment variables TMPDIR, TMP, or TEMP, or to the directory /tmp.

The file is then opened using mode “w+”, which allows reading and writing and deletes any existing content (see Table 22.5).

open

Tempfile.open( basename, tmpdir ) → ref

Synonym for Tempfile.new.

Tempfile instance methods

close: ref.close( final=false ); Closes ref. If final is true, deletes the underlying real file. If final is false, ref may be subsequently reopened. In all cases, the underlying file is deleted when the program terminates.
open: ref.open; Reopens ref using mode “r+”, which allows reading and writing but does not delete existing content.
path: ref.path → aString; Returns the full path of the underlying file.

class Mutex < Object

require 'thread'
sema4 = Mutex.new

a = Thread.new {
  sema4.synchronize {
    # access shared resource
  }
}

b = Thread.new {
  sema4.synchronize {
    # access shared resource
  }
}

Mutex implements a simple semaphore that can be used to coordinate access to shared data from multiple concurrent threads.

Mutex instance methods

lock: ref.lock → ref; Attempts to grab the lock and waits if it isn't available.
locked?: ref.locked? → true or false; Returns true if this lock is currently held by some thread.
synchronize: ref.synchronize { block } → ref; Obtains a lock (using Mutex#lock), runs the block, and releases the lock when the block completes.
try_lock: ref.try_lock → true or false; Attempts to obtain the lock and returns immediately. Returns true if the lock was granted.
unlock: ref.unlock → ref or nil; Releases the lock. Returns nil if ref wasn't locked.

class ConditionVariable < Object

require 'thread'
mutex = Mutex.new
resource = ConditionVariable.new

a = Thread.new {
  mutex.synchronize {
    # Thread 'a' now needs the resource
    resource.wait(mutex)
    # 'a' can now have the resource
  }
}

b = Thread.new {
  mutex.synchronize {
    # Thread 'b' has finished using the resource
    resource.signal
  }
}

ConditionVariable objects augment class Mutex. Using condition variables, it is possible to suspend while in the middle of a critical section until a resource becomes available (see the discussion “Condition Variables”).

ConditionVariable instance methods

broadcast: ref.broadcast; Wakes up all threads waiting for this lock.
signal: ref.signal; Wakes up the first thread in line waiting for this lock.
wait: ref.wait( aMutex ) → aMutex; Releases the lock held in aMutex and waits; reacquires the lock on wakeup.

library timeout

require "timeout"

for snooze in 1..2
  puts "About to sleep for #{snooze}"
  timeout(1.5) do
    sleep(snooze)
  end
  puts "That was refreshing"
end

produces:

About to sleep for 1
That was refreshing
About to sleep for 2
/tc/usr/lib/ruby/1.6/timeout.rb:37: execution expired (TimeoutError)
	from prog.rb:5:in `timeout'
	from prog.rb:5
	from prog.rb:3:in `each'
	from prog.rb:3

The timeout method takes a single parameter, representing a timeout period in seconds, and a block. The block is executed, and a timer is run concurrently. If the block terminates before the timeout, timeout returns true. Otherwise, a TimeoutError exception is raised.

class WeakRef < Delegator

require "weakref"

ref = "fol de rol"
puts "Initial object is #{ref}"
ref = WeakRef.new(ref)
puts "Weak reference is #{ref}"
ObjectSpace.garbage_collect
puts "But then it is #{ref}"

produces:

Initial object is fol de rol
Weak reference is fol de rol
prog.rb:8: Illegal Reference - probably recycled (WeakRef::RefError)

In Ruby, objects are not eligible for garbage collection if there are still references to them. Normally, this is a Good Thing—it would be disconcerting to have an object simply evaporate while you were using it. However, sometimes you may need more flexibility. For example, you might want to implement an in-memory cache of commonly used file contents. As you read more files, the cache grows. At some point, you may run low on memory. The garbage collector will be invoked, but the objects in the cache are all referenced by the cache data structures, and so will not be deleted.

A weak reference behaves exactly as any normal object reference with one important exception—the referenced object may be garbage collected, even while references to it exist. In the cache example, if the cached files were accessed using weak references, once memory runs low they will be garbage collected, freeing memory for the rest of the application.

Weak references introduce a slight complexity. As the object referenced can be deleted by garbage collection at any time, code that accesses these objects must take care to ensure that the references are valid. Two techniques can be used. First, the code can reference the objects normally. Any attempt to reference an object that has been garbage collected will raise a WeakRef::RefError exception.

An alternative approach is to use the WeakRef#weakref_alive? method to check that a reference is valid before using it. Garbage collection must be disabled during the test and subsequent reference to the object. In a single-threaded program, you could use something like:

ref = WeakRef.new(someObject)
#
# .. some time later
#

gcWasDisabled = GC.disable
if ref.weakref_alive?
  # do stuff with 'ref'
end
GC.enable unless gcWasDisabled

WeakRef class methods

new: WeakRef.new( anObject ) → ref; Creates and returns a weak reference to anObject. All future references to anObject should be made using ref.

WeakRef instance methods

weakref_alive?: ref.weakref_alive? → true or false; Returns false if the object referenced by ref has been garbage collected.

Extracted from the book "Programming Ruby - The Pragmatic Programmer's Guide"

Copyright © 2001 by Addison Wesley Longman, Inc. This material may be distributed only subject to the terms and conditions set forth in the Open Publication License, v1.0 or later (the latest version is presently available at http://www.opencontent.org/openpub/).

Distribution of substantively modified versions of this document is prohibited without the explicit permission of the copyright holder.

Distribution of the work or derivative of the work in any standard (paper) book form is prohibited unless prior permission is obtained from the copyright holder.