From: Ian Jackson <ijackson@chiark.greenend.org.uk>
Date: Sun, 21 Aug 2016 09:38:15 +0000 (+0100)
Subject: New format: README.format and normalise, wip.  Before restrict CAND and VOTERNAME... 
X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ian/git?p=appendix-a6.git;a=commitdiff_plain;h=a11536b183032286d08363afa8f0d1fbd124b979;ds=sidebyside

New format: README.format and normalise, wip.  Before restrict CAND and VOTERNAME to \w+
---

diff --git a/README.format b/README.format
new file mode 100644
index 0000000..e08627c
--- /dev/null
+++ b/README.format
@@ -0,0 +1,67 @@
+We define a common election and ballot input format.
+
+Format consists of lines:
+
+  [| OPTION]
+  CAND = DESCRIPTION ... [| OPTION...]
+  VOTERNAME : CAND, (CAND=CAND=...) ... [| OPTION...]
+
+
+We define normalised and non-normalised inputs.  Normalised is for
+input by compute programs (which do not need to check conformance with
+the syntax) and non-normalised is for output from converters and
+messing about by numans.
+
+In normalised files lines must occor in the order above (and
+whitespace is normalised, although readers should avoid relying on
+this other than that they may assume lack of leading and trailing
+whitespace on each line).
+
+Non-normalised files can also contain empty lines and
+  # COMMENT...
+
+In non-normalised files any CAND may be omitted in which case CAND =
+CAND is assumed.  In non-normalised files an empty description is
+replaced with a copy of CAND.
+
+CAND and VOTERNAME may contain any characters except control
+characters and 7-bit ASCII punctuation, but including ASCII
+  ampersand &
+  apostrophe '
+  hyphen -
+  full stop .
+  and underscore (_).  Processors are not required to
+generate unambiguous output when CAND or VOTERNAME contains characters
+other than 7-bit alphanumerics plus _.
+
+DESCRIPTION may contain any characters except | (and cannot contain
+leading or trailing whitespace).
+
+VOTERNAMEs need not be distinct and may be empty.
+
+OPTION is OPTNAME[=VALUE] and modifies the preference, ballot, or whole
+election.  VALUE may contain no whitespace.
+
+OPTNAME contains only alphanumerics (or _) and starts with a letter or
+_.  Reader should treat unknown OPTNAMEs as follows:
+  starts with uppercase letter - crash (reject the input file)
+  starts with lowercase letter - generate a warning
+  starts with underscore - silently ignore
+
+Anything which expects a particular OPTNAME should usually accept all
+three variants, with a regexp like /^_?[Qq]orum=(\d+)$/.  Convention
+is _-separated words, with inner words being in lowercase, even when
+OPTNAME starts with a capital.
+
+Currently defined OPTNAMEs (default values shown)
+
+ Election   seats=1     Number of seats to be filled.
+ Election   _nodefault  For Debian A6, no default option
+ Election   quorum=NUM  Quorum (default, none)
+
+ Candidate  Super=1:1   For Debian A6, RAT:IO, 
+ Candidate  default     For Debian A6, Is the default option
+
+ Voter      Weight=1.0
+
+OPTIONs wihin a particular scope may occur in any order.
diff --git a/normalise b/normalise
new file mode 100755
index 0000000..ca5ecf7
--- /dev/null
+++ b/normalise
@@ -0,0 +1,40 @@
+#!/usr/bin/perl -w
+
+use strict;
+
+our @options, @candiates, @voters;
+
+our %seen_cand, %need_cand;
+
+my $candvoter_re = '[^\000-\037!"#$%()*+,/0-\136`-\177]+';
+
+sub normalise_opts ($) {
+    my ($os) = @_;
+    my @o;
+    foreach my $o (split /\s+/, $os) {
+	if ($o =~ m/^\w+$/) {
+	    push @o, $&;
+	} elsif ($o =~ m/^\w+\=\S+$/) {
+	    push @o, $&;S
+	} elseif ($o !~ m/\S/) {
+	} else {
+	    badinput "bad option \`$o'";
+	}
+    }
+    return @o ? " | @o" : "";
+}
+
+while (<>) {
+    next unless m/\S/;
+    next if m/^\#/;
+    s/^\s+//;
+    s/\s+$//;
+    if (m/^\|\s*(\w+(?:\=\S+)?)$/) {
+	push @options, "| $1";
+    } elsif (m/^($candvoter_re?)\s*=\s*([^|]+?)\s*|(.*)?$/) {
+	my ($cand,$desc,$opts) = ($1,$2,$3);
+	$desc=$cand unless length $desc;
+	$opts = normalise_opts $opts;
+	push @candidates, "$cand = $desc".$opts;
+    } elsif (m/^($candvoter_re?)?\s*\:/) {
+