[PATCH] lib: Fix RegexpPostingSource

[David Bremner]

From: Olly Betts <olly at survex.com>

Remove incorrect skipping to first match from init(), and add explicit
skip_to() and check() methods to work around xapian-core bug (the
check() method will also improve speed when filtering by one of
these).
---

I was reading the Xapian API docs last night (as one does when home
with a cold) and realized that the last three lines of
RegexpPostingSource::init should not be necessary according to the
docs.  I asked Olly about it, and he very kindly dug into the notmuch
sources and provided this improved workaround. This should go in 0.24.

 lib/regexp-fields.cc | 26 +++++++++++++++++++++-----
 lib/regexp-fields.h  |  2 ++
 2 files changed, 23 insertions(+), 5 deletions(-)

diff --git a/lib/regexp-fields.cc b/lib/regexp-fields.cc
index b4174750..8e740a81 100644
--- a/lib/regexp-fields.cc
+++ b/lib/regexp-fields.cc
@@ -62,11 +62,6 @@ RegexpPostingSource::init (const Xapian::Database &db)
     it_ = db_.valuestream_begin (slot_);
     end_ = db.valuestream_end (slot_);
     started_ = false;
-
-    /* make sure we start on a matching value */
-    while (!at_end() && regexec (&regexp_, (*it_).c_str (), 0, NULL, 0) != 0) {
-	++it_;
-    }
 }
 
 Xapian::doccount
@@ -113,6 +108,27 @@ RegexpPostingSource::next (unused (double min_wt))
     }
 }
 
+void
+RegexpPostingSource::skip_to (Xapian::docid did, unused (double min_wt))
+{
+    started_ = true;
+    it_.skip_to (did);
+    for (; ! at_end (); ++it_) {
+	std::string value = *it_;
+	if (regexec (&regexp_, value.c_str (), 0, NULL, 0) == 0)
+	    break;
+    }
+}
+
+bool
+RegexpPostingSource::check (Xapian::docid did, unused (double min_wt))
+{
+    started_ = true;
+    if (!it_.check (did) || at_end ())
+	return false;
+    return (regexec (&regexp_, (*it_).c_str (), 0, NULL, 0) == 0);
+}
+
 static inline Xapian::valueno _find_slot (std::string prefix)
 {
     if (prefix == "from")
diff --git a/lib/regexp-fields.h b/lib/regexp-fields.h
index bac11999..a4ba7ad8 100644
--- a/lib/regexp-fields.h
+++ b/lib/regexp-fields.h
@@ -56,6 +56,8 @@ class RegexpPostingSource : public Xapian::PostingSource
     Xapian::docid get_docid () const;
     bool at_end () const;
     void next (unused (double min_wt));
+    void skip_to (Xapian::docid did, unused (double min_wt));
+    bool check (Xapian::docid did, unused (double min_wt));
 };
 
 
-- 
2.11.0