Changes to m4/doc/m4.texinfo,v [branch-1_4]

[Eric Blake]

CVSROOT:        /sources/m4
Module name:    m4
Branch:         branch-1_4
Changes by:     Eric Blake <ericb>      06/10/21 02:55:56

Index: doc/m4.texinfo
===================================================================
RCS file: /sources/m4/m4/doc/m4.texinfo,v
retrieving revision 1.1.1.1.2.90
retrieving revision 1.1.1.1.2.91
diff -u -b -r1.1.1.1.2.90 -r1.1.1.1.2.91
--- doc/m4.texinfo      19 Oct 2006 23:13:05 -0000      1.1.1.1.2.90
+++ doc/m4.texinfo      21 Oct 2006 02:55:56 -0000      1.1.1.1.2.91
@@ -1629,7 +1629,7 @@
 ')
 @result{}
 string
address@hidden macro @comment
address@hidden address@hidden }
 defn(`string')
 @result{}The macro dnl is very useful
 @result{}
@@ -2187,7 +2187,7 @@
 include(`forloop.m4')
 @result{}
 forloop(`i', `1', `8', `i ')
address@hidden 2 3 4 5 6 7 8 @comment
address@hidden 2 3 4 5 6 7 address@hidden }
 @end example
 
 For-loops can be nested, like:
@@ -2207,12 +2207,12 @@
 The implementation of the @code{forloop} macro is fairly
 straightforward.  The @code{forloop} macro itself is simply a wrapper,
 which saves the previous definition of the first argument, calls the
-internal macro @code{_forloop}, and re-establishes the saved definition of
-the first argument.
+internal macro @address@hidden, and re-establishes the saved
+definition of the first argument.
 
-The macro @code{_forloop} expands the fourth argument once, and tests
-to see if the iterator has reached the final value.  If it has not
-finished, it increments the iterator (using the predefined macro
+The macro @address@hidden expands the fourth argument once, and
+tests to see if the iterator has reached the final value.  If it has
+not finished, it increments the iterator (using the predefined macro
 @code{incr}, @pxref{Incr}), and recurses.
 
 Here is an actual implementation of @code{forloop}, distributed as
@@ -2240,7 +2240,151 @@
 @node Foreach
 @section Iteration by list contents
 
-FIXME - Fill out this section.
address@hidden for each loops
address@hidden loops, list iteration
address@hidden iterating over lists
+Here is an example of a loop macro that implements list iteration.
+
address@hidden Composite foreach (@var{iterator}, @var{paren-list}, @var{text})
address@hidden Composite foreachq (@var{iterator}, @var{quote-list}, @var{text})
+Takes the name in @var{iterator}, which must be a valid macro name, and
+successively assign it each value from @var{paren-list} or
address@hidden  In @code{foreach}, @var{paren-list} is a
+comma-separated list of elements contained in parentheses.  In
address@hidden, @var{quote-list} is a comma-separated list of elements
+contained in a quoted string.  For each assignment to @var{iterator},
+append @var{text} to the overall expansion.  @var{text} may refer to
address@hidden  Any definition of @var{iterator} prior to this
+invocation is restored.
address@hidden deffn
+
+As an example, this displays each word in a list inside of a sentence,
+using an implementation of @code{foreach} distributed as
address@hidden@value{VERSION}/@/examples/@/foreach.m4}, and @code{foreachq}
+in @address@hidden/@/examples/@/foreachq.m4}.
+
address@hidden
+include(`foreach.m4')
address@hidden
+foreach(`x', (foo, bar, foobar), `Word was: x
+')dnl
address@hidden was: foo
address@hidden was: bar
address@hidden was: foobar
+include(`foreachq.m4')
address@hidden
+foreachq(`x', `foo, bar, foobar', `Word was: x
+')dnl
address@hidden was: foo
address@hidden was: bar
address@hidden was: foobar
address@hidden example
+
+It is possible to be more complex; each element of the @var{paren-list}
+or @var{quote-list} can itself be a list, to pass as further arguments
+to a helper macro.  This example generates a shell case statement:
+
address@hidden
+include(`foreach.m4')
address@hidden
+define(`_case', `  $1)
+    $2=" $1";;
+')dnl
+define(`_cat', `$1$2')dnl
+case $`'1 in
address@hidden $1 in
+foreach(`x', `(`(`a', `vara')', `(`b', `varb')', `(`c', `varc')')',
+        `_cat(`_case', x)')dnl
address@hidden  a)
address@hidden    vara=" a";;
address@hidden  b)
address@hidden    varb=" b";;
address@hidden  c)
address@hidden    varc=" c";;
+esac
address@hidden
address@hidden example
+
+The implementation of the @code{foreach} macro is a bit more involved;
+it is a wrapper around two helper macros.  First, @address@hidden is
+needed to grab the first element of a list.  Second,
address@hidden@w{_foreach}} implements the recursion, successively walking
+through the original list.  Here is a simple implementation of
address@hidden:
+
address@hidden
+undivert(`foreach.m4')dnl
address@hidden(`-1')
address@hidden foreach(x, (item_1, item_2, ..., item_n), stmt)
address@hidden   parenthesized list, simple version
address@hidden(`foreach', `pushdef(`$1')_foreach($@@)popdef(`$1')')
address@hidden(`_arg1', `$1')
address@hidden(`_foreach', `ifelse(`$2', `()', `',
address@hidden  `define(`$1', _arg1$2)$3`'$0(`$1', (shift$2), `$3')')')
address@hidden'dnl
address@hidden example
+
+Unfortunately, that implementation is not robust to macro names as list
+elements.  Each iteration of @address@hidden is stripping another
+layer of quotes, leading to erratic results if list elements are not
+already fully expanded.  The first cut at implementing @code{foreachq}
+takes this into account.  Also, when using quoted elements in a
address@hidden, the overall list must be quoted.  A @var{quote-list}
+has the nice property of requiring fewer characters to create a list
+containing the same quoted elements.  To see the difference between the
+two macros, we attempt to pass double-quoted macro names in a list,
+expecting the macro name on output after one layer of quotes is removed
+during list iteration and the final layer removed during the final
+rescan:
+
address@hidden
+define(`a', `1')define(`b', `2')define(`c', `3')
address@hidden
+include(`foreach.m4')
address@hidden
+include(`foreachq.m4')
address@hidden
+foreach(`x', `(``a'', ``(b'', ``c)'')', `x
+')
address@hidden
address@hidden(2)1
address@hidden
address@hidden, x
address@hidden)
+foreachq(`x', ```a'', ``(b'', ``c)''', `x
+')dnl
address@hidden
address@hidden(b
address@hidden)
address@hidden example
+
+Obviously, @code{foreachq} did a better job; here is its implementation:
+
address@hidden
+undivert(`foreachq.m4')dnl
address@hidden(`quote.m4')dnl
address@hidden(`-1')
address@hidden foreachq(x, `item_1, item_2, ..., item_n', stmt)
address@hidden   quoted list, simple version
address@hidden(`foreachq', `pushdef(`$1')_foreachq($@@)popdef(`$1')')
address@hidden(`_arg1', `$1')
address@hidden(`_foreachq', `ifelse(quote($2), `', `',
address@hidden  `define(`$1', `_arg1($2)')$3`'$0(`$1', `shift($2)', `$3')')')
address@hidden'dnl
address@hidden example
+
+Notice that @address@hidden had to use the helper macro
address@hidden defined earlier (@pxref{Shift}), to ensure that the
+embedded @code{ifelse} call does not go haywire if a list element
+contains a comma.  Unfortunately, this implementation of @code{foreachq}
+has its own severe flaw.  Whereas the @code{foreach} implementation was
+linear, this macro is quadratic in the number of list elements, and is
+much more likely to trip up the limit set by the command line option
address@hidden (or @option{-L}, @pxref{Limits control, ,
+Invoking m4}).  (It is possible to have robust iteration with linear
+behavior for either list style.  See if you can learn from the best
+elements of both of these implementations to create robust macros; or
address@hidden foreach, , Answers}).
 
 @node Debugging
 @chapter How to debug macros and input
@@ -3937,7 +4081,7 @@
 patsubst(`GNUs not Unix', `\w+', `(\&)')
 @result{}(GNUs) (not) (Unix)
 patsubst(`GNUs not Unix', `[A-Z][a-z]+')
address@hidden not @comment
address@hidden address@hidden }
 patsubst(`GNUs not Unix', `not', `NOT\')
 @error{}m4:stdin:6: Warning: trailing \ ignored in replacement
 @result{}GNUs NOT Unix
@@ -4307,10 +4451,10 @@
 string.
 @end deffn
 
-When @acronym{GNU} extensions are in effect (that is, when you did not use the
address@hidden option, @pxref{Limits control, , Invoking m4}),
address@hidden @code{m4} will
-define the macro @code{__gnu__} to expand to the empty string.
+When @acronym{GNU} extensions are in effect (that is, when you did not
+use the @option{-G} option, @pxref{Limits control, , Invoking m4}),
address@hidden @code{m4} will define the macro @address@hidden to
+expand to the empty string.
 
 @example
 __gnu__
@@ -4320,15 +4464,15 @@
 @end example
 
 @cindex platform macro
-On UNIX systems, @acronym{GNU} @code{m4} will define @code{__unix__} by
-default, or @code{unix} when the @option{-G} option is specified.
+On UNIX systems, @acronym{GNU} @code{m4} will define @address@hidden
+by default, or @code{unix} when the @option{-G} option is specified.
 
 On native Windows systems, @acronym{GNU} @code{m4} will define
address@hidden by default, or @code{windows} when the @option{-G}
-option is specified.
address@hidden@w{__windows__}} by default, or @code{windows} when the
address@hidden option is specified.
 
-On OS/2 systems, @acronym{GNU} @code{m4} will define @code{__os2__} by
-default, or @code{os2} when the @option{-G} option is specified.
+On OS/2 systems, @acronym{GNU} @code{m4} will define @address@hidden
+by default, or @code{os2} when the @option{-G} option is specified.
 
 If @acronym{GNU} @code{m4} does not provide a platform macro for your system,
 please report that as a bug.
@@ -4719,16 +4863,16 @@
 @result{}6
 @end example
 
-The @code{__program__} macro behaves like @samp{$0} in shell
+The @address@hidden macro behaves like @samp{$0} in shell
 terminology.  If you invoke @code{m4} through an absolute path or a link
 with a different spelling, rather than by relying on a @env{PATH} search
-for plain @samp{m4}, it will affect how @code{__program__} expands.  The
-intent is that you can use it to produce error messages with the same
-formatting that @code{m4} produces internally.  It can also be used
+for plain @samp{m4}, it will affect how @address@hidden expands.
+The intent is that you can use it to produce error messages with the
+same formatting that @code{m4} produces internally.  It can also be used
 within @code{syscmd} (@pxref{Syscmd}) to pick the same version of
 @code{m4} that is currently running, rather than whatever version of
address@hidden happens to be first in @env{PATH}.  It was first introduced
-in @acronym{GNU} M4 1.4.6.
address@hidden happens to be first in @env{PATH}.  It was first introduced in
address@hidden M4 1.4.6.
 
 @node M4exit
 @section Exiting from @code{m4}
@@ -5069,8 +5213,8 @@
 
 @item
 The name of the program, the current input file, and the current input
-line number are accessible through the builtins @code{__program__},
address@hidden, and @code{__line__} (@pxref{Location}).
+line number are accessible through the builtins @address@hidden,
address@hidden@w{__file__}}, and @address@hidden (@pxref{Location}).
 
 @item
 The format of the output from @code{dumpdef} and macro tracing can be
@@ -5342,7 +5486,7 @@
 only permits decimal numbers for bounds.  Here is an improved version,
 shipped as @address@hidden/@/examples/@/forloop2.m4}; this
 version also optimizes based on the fact that the starting bound does
-not need to be passed to the helper @code{_forloop}.
+not need to be passed to the helper @address@hidden
 
 @example
 undivert(`forloop2.m4')dnl
@@ -5379,7 +5523,199 @@
 @node Improved foreach
 @section Solution for @code{foreach}
 
-FIXME - add content.
+The @code{foreach} and @code{foreachq} macros (@pxref{Foreach}) as
+presented earlier each have flaws.  First, we will examine and fix the
+quadratic behavior of @code{foreachq}:
+
address@hidden
+include(`foreachq.m4')
address@hidden
+traceon(`shift')debugmode(`aq')
address@hidden
+foreachq(`x', ``1', `2', `3', `4'', `x
+')dnl
address@hidden
address@hidden: -3- shift(`1', `2', `3', `4')
address@hidden: -2- shift(`1', `2', `3', `4')
address@hidden
address@hidden: -4- shift(`1', `2', `3', `4')
address@hidden: -3- shift(`2', `3', `4')
address@hidden: -3- shift(`1', `2', `3', `4')
address@hidden: -2- shift(`2', `3', `4')
address@hidden
address@hidden: -5- shift(`1', `2', `3', `4')
address@hidden: -4- shift(`2', `3', `4')
address@hidden: -3- shift(`3', `4')
address@hidden: -4- shift(`1', `2', `3', `4')
address@hidden: -3- shift(`2', `3', `4')
address@hidden: -2- shift(`3', `4')
address@hidden
address@hidden: -6- shift(`1', `2', `3', `4')
address@hidden: -5- shift(`2', `3', `4')
address@hidden: -4- shift(`3', `4')
address@hidden: -3- shift(`4')
address@hidden example
+
+Each successive iteration was adding more quoted @code{shift}
+invocations, and the entire list contents were passing through every
+iteration.  In general, when recursing, it is a good idea to make the
+recursion use fewer arguments, rather than adding additional quoted
+uses of @code{shift}.  By doing so, @code{m4} uses less memory, invokes
+fewer macros, is less likely to run into machine limits, and most
+importantly, performs faster.  The fixed version of @code{foreachq} can
+be found in @address@hidden/@/examples/@/foreachq2.m4}:
+
address@hidden
+include(`foreachq2.m4')
address@hidden
+undivert(`foreachq2.m4')dnl
address@hidden(`quote.m4')dnl
address@hidden(`-1')
address@hidden foreachq(x, `item_1, item_2, ..., item_n', stmt)
address@hidden   quoted list, improved version
address@hidden(`foreachq', `pushdef(`$1')_foreachq($@@)popdef(`$1')')
address@hidden(`_arg1q', ``$1'')
address@hidden(`_rest', `ifelse(`$#', `1', `', `dquote(shift($@@))')')
address@hidden(`_foreachq', `ifelse(`$2', `', `',
address@hidden  `define(`$1', _arg1q($2))$3`'$0(`$1', _rest($2), `$3')')')
address@hidden'dnl
+traceon(`shift')debugmode(`aq')
address@hidden
+foreachq(`x', ``1', `2', `3', `4'', `x
+')dnl
address@hidden
address@hidden: -3- shift(`1', `2', `3', `4')
address@hidden
address@hidden: -3- shift(`2', `3', `4')
address@hidden
address@hidden: -3- shift(`3', `4')
address@hidden
address@hidden example
+
+Note that the fixed version calls unquoted helper macros in
address@hidden@w{_foreachq}} to trim elements immediately; those helper macros
+in turn must re-supply the layer of quotes lost in the macro invocation.
+Contrast the use of @address@hidden, which quotes the first list
+element, with @address@hidden of the earlier implementation that
+returned the first list element directly.
+
+For a different approach, the improved version of @code{foreach},
+available in @address@hidden/@/examples/@/foreach2.m4}, simply
+overquotes the arguments to @address@hidden to begin with, using
address@hidden  Then @address@hidden can just use
address@hidden@w{_arg1}} to remove the extra layer of quoting that was added up
+front:
+
address@hidden
+include(`foreach2.m4')
address@hidden
+undivert(`foreach2.m4')dnl
address@hidden(`quote.m4')dnl
address@hidden(`-1')
address@hidden foreach(x, (item_1, item_2, ..., item_n), stmt)
address@hidden   parenthesized list, improved version
address@hidden(`foreach', `pushdef(`$1')_foreach(`$1',
address@hidden  (dquote(dquote_elt$2)), `$3')popdef(`$1')')
address@hidden(`_arg1', `$1')
address@hidden(`_foreach', `ifelse(`$2', `(`')', `',
address@hidden  `define(`$1', _arg1$2)$3`'$0(`$1', (dquote(shift$2)), `$3')')')
address@hidden'dnl
+traceon(`shift')debugmode(`aq')
address@hidden
+foreach(`x', `(`1', `2', `3', `4')', `x
+')dnl
address@hidden: -4- shift(`1', `2', `3', `4')
address@hidden: -4- shift(`2', `3', `4')
address@hidden: -4- shift(`3', `4')
address@hidden
address@hidden: -3- shift(``1'', ``2'', ``3'', ``4'')
address@hidden
address@hidden: -3- shift(``2'', ``3'', ``4'')
address@hidden
address@hidden: -3- shift(``3'', ``4'')
address@hidden
address@hidden: -3- shift(``4'')
address@hidden example
+
+In summary, recursion over list elements is trickier than it appeared at
+first glance, but provides a powerful idiom within @code{m4} processing.
+As a final demonstration, both list styles are now able to handle
+several scenarios that would wreak havoc on the original
+implementations.  This points out one other difference between the two
+list styles.  @code{foreach} evaluates unquoted list elements only once,
+in preparation for calling @address@hidden  But @code{foreachq}
+evaluates unquoted list elements twice while visiting the first list
+element, once in @address@hidden and once in @address@hidden  When
+deciding which list style to use, one must take into account whether
+repeating the side effects of unquoted list elements will have any
+detrimental effects.
+
address@hidden
+include(`foreach2.m4')
address@hidden
+include(`foreachq2.m4')
address@hidden
+dnl 0-element list:
+foreach(`x', `', `<x>') / foreachq(`x', `', `<x>')
address@hidden /@w{ }
+dnl 1-element list of empty element
+foreach(`x', `()', `<x>') / foreachq(`x', ``'', `<x>')
address@hidden<> / <>
+dnl 2-element list of empty elements
+foreach(`x', `(`',`')', `<x>') / foreachq(`x', ``',`'', `<x>')
address@hidden<><> / <><>
+dnl 1-element list of a comma
+foreach(`x', `(`,')', `<x>') / foreachq(`x', ``,'', `<x>')
address@hidden<,> / <,>
+dnl 2-element list of unbalanced parentheses
+foreach(`x', `(`(', `)')', `<x>') / foreachq(`x', ``(', `)'', `<x>')
address@hidden<(><)> / <(><)>
+define(`active', `ACT, IVE')
address@hidden
+traceon(`active')
address@hidden
+dnl list of unquoted macros; expansion occurs before recursion
+foreach(`x', `(active, active)', `<x>
+')dnl
address@hidden: -4- active -> `ACT, IVE'
address@hidden: -4- active -> `ACT, IVE'
address@hidden<ACT>
address@hidden<IVE>
address@hidden<ACT>
address@hidden<IVE>
+foreachq(`x', `active, active', `<x>
+')dnl
address@hidden: -3- active -> `ACT, IVE'
address@hidden: -3- active -> `ACT, IVE'
address@hidden<ACT>
address@hidden: -3- active -> `ACT, IVE'
address@hidden: -3- active -> `ACT, IVE'
address@hidden<IVE>
address@hidden<ACT>
address@hidden<IVE>
+dnl list of quoted macros; expansion occurs during recursion
+foreach(`x', `(`active', `active')', `<x>
+')dnl
address@hidden: -1- active -> `ACT, IVE'
address@hidden<ACT, IVE>
address@hidden: -1- active -> `ACT, IVE'
address@hidden<ACT, IVE>
+foreachq(`x', ``active', `active'', `<x>
+')dnl
address@hidden: -1- active -> `ACT, IVE'
address@hidden<ACT, IVE>
address@hidden: -1- active -> `ACT, IVE'
address@hidden<ACT, IVE>
+dnl list of double-quoted macro names; no expansion
+foreach(`x', `(``active'', ``active'')', `<x>
+')dnl
address@hidden<active>
address@hidden<active>
+foreachq(`x', ```active'', ``active''', `<x>
+')dnl
address@hidden<active>
address@hidden<active>
address@hidden example
 
 @node Improved cleardivert
 @section Solution for @code{cleardivert}
@@ -5419,12 +5755,13 @@
 @section Solution for @code{fatal_error}
 
 The @code{fatal_error} macro (@pxref{M4exit}) is not robust to versions
-of @acronym{GNU} M4 earlier than 1.4.8, where invoking @code{__file__}
-(@pxref{Location}) inside @code{m4wrap} would result in an empty string,
-and @code{__line__} resulted in @samp{0} even though all files start at
-line 1.  Furthermore, versions earlier than 1.4.6 did not support the
address@hidden macro.  If you want @code{fatal_error} to work across
-the entire 1.4.x release series, a better implementation would be:
+of @acronym{GNU} M4 earlier than 1.4.8, where invoking
address@hidden@w{__file__}} (@pxref{Location}) inside @code{m4wrap} would result
+in an empty string, and @address@hidden resulted in @samp{0} even
+though all files start at line 1.  Furthermore, versions earlier than
+1.4.6 did not support the @address@hidden macro.  If you want
address@hidden to work across the entire 1.4.x release series, a
+better implementation would be:
 
 @example
 define(`fatal_error',