new module extension

1 files changed, 92 insertions, 0 deletions

diff --git a/lib/MaildirIndexer/Index/ByAddresses.rakumod b/lib/MaildirIndexer/Index/ByAddresses.rakumod
new file mode 100644
index 0000000..b83a239
--- /dev/null
+++ b/lib/MaildirIndexer/Index/ByAddresses.rakumod
@@ -0,0 +1,92 @@
+use v6.d;
+use MaildirIndexer::Index;
+unit class MaildirIndexer::Index::ByAddresses does MaildirIndexer::Index;
+use MaildirIndexer::LogTimelineSchema;
+use MaildirIndexer::Email;
+
+# most of this is copied from
+# p6-Algorithm-NaiveBayes:auth<cpan:TITSUKI>:ver<0.04>, in particular the
+# Algorithm::NaiveBayes::Classifier::Bernoulli class
+
+has Array[Str] %!addresses-for-file;
+# I'd like to type-constrain these BagHash-es, but the compiler
+# currently dies if I try
+has BagHash $!count-by-address-and-mailbox .= new;
+has BagHash $!known-addresses .= new;
+has BagHash $!count-by-mailbox .= new;
+has $!total-count;
+
+method dump() {
+}
+
+submethod account-for(Str @addresses,Str $mailbox,Int $step) {
+    $!total-count += $step;
+    $!count-by-mailbox{$mailbox} += $step;
+
+    for @addresses -> Str $addr {
+        $!known-addresses{$addr} += $step;
+        $!count-by-address-and-mailbox{$addr => $mailbox} += $step;
+    }
+}
+
+method add-mail(MaildirIndexer::Email:D $email, Str:D $mailbox --> Nil) {
+    MaildirIndexer::LogTimelineSchema::Index::Add.log: :class('ByAddresses'),:$mailbox, -> {
+        my Str @addresses = $email.addresses or return;
+        %!addresses-for-file{ $email.path } = @addresses;
+
+        self.account-for(@addresses,$mailbox,1);
+
+        return;
+    }
+}
+
+method del-path(IO:D $file, Str:D $mailbox --> Nil) {
+    MaildirIndexer::LogTimelineSchema::Index::Rm.log: :class('ByAddresses'),:$mailbox, -> {
+        my Str @addresses = %!addresses-for-file{$file.path} or return;
+
+        self.account-for(@addresses,$mailbox,-1);
+
+        return;
+    }
+}
+
+submethod p-address-given-mailbox(Str $addr, Str $mailbox) {
+    my $a = 1e-15 + $!count-by-address-and-mailbox{$addr => $mailbox};
+    my $b = 1 + $!count-by-mailbox{$mailbox};
+    return $a / $b;
+}
+
+submethod predict-mailbox-given-addresses(@addresses) {
+    my %prediction;
+    my Bag $addr-bag .= new(|@addresses);
+
+    for $!count-by-mailbox.keys -> Str $mailbox {
+        my $p = $!count-by-mailbox{$mailbox} / $!total-count;
+
+        for $!known-addresses.keys -> Str $addr {
+            my $addr-p = self.p-address-given-mailbox($addr,$mailbox);
+            if ($addr-bag{$addr}) {
+                $p *= $addr-p;
+            }
+            else {
+                $p *= 1 - $addr-p;
+            }
+        }
+
+        %prediction{$mailbox} = $p;
+    }
+
+    return %prediction;
+}
+
+method mailbox-for-email(MaildirIndexer::Email:D $email --> Str) {
+    my Str $result;
+    MaildirIndexer::LogTimelineSchema::Index::Find.log: :class('ByAddresses'), -> {
+        my %prediction = self.predict-mailbox-given-addresses($email.addresses);
+
+        my @most-probable-mailboxes = %prediction.pairs.sort(*.value);
+
+        if @most-probable-mailboxes -> $_ { $result = .[*-1].key }
+    }
+    return $result;
+}