[PATCH] lib: add 'body:' field, stop indexing headers twice.

[David Bremner <david@tethera.net>]

The new `body:` field (in Xapian terms) or prefix (in slightly
sloppier notmuch) terms allows matching terms that occur only in the
body.

Unprefixed query terms should continue to match anywhere (header or
body) in the message.

This follows a suggestion of Olly Betts to use the facility (since
Xapian 1.0.4) to add the same field with multiple prefixes. The double
indexing of previous versions is thus replaced with a query time
expension of unprefixed query terms to the various prefixed
equivalent.

Reindexing will be needed for 'body:' searches to work correctly;
otherwise they will also match messages where the term occur in
headers (demonstrated by the new tests in T530-upgrade.sh)
---

Compared to the previous version this adds a couple more tests and
clarifies the commit message comment about reindexing.

 doc/man7/notmuch-search-terms.rst |  5 +++-
 lib/database-private.h            |  6 +++++
 lib/database.cc                   | 20 +++++++++++---
 lib/message.cc                    | 22 +++++++---------
 test/T530-upgrade.sh              | 16 ++++++++++++
 test/T730-body.sh                 | 43 +++++++++++++++++++++++++++++++
 6 files changed, 95 insertions(+), 17 deletions(-)
 create mode 100755 test/T730-body.sh

diff --git a/doc/man7/notmuch-search-terms.rst b/doc/man7/notmuch-search-terms.rst
index f7a39ceb..fd8bf634 100644
--- a/doc/man7/notmuch-search-terms.rst
+++ b/doc/man7/notmuch-search-terms.rst
@@ -44,6 +44,9 @@ results to those whose value matches a regular expression (see
 
    notmuch search 'from:"/bob@.*[.]example[.]com/"'
 
+body:<word-or-quoted-phrase>
+    Match terms in the body of messages.
+
 from:<name-or-address> or from:/<regex>/
     The **from:** prefix is used to match the name or address of
     the sender of an email message.
@@ -249,7 +252,7 @@ follows.
 Boolean
    **tag:**, **id:**, **thread:**, **folder:**, **path:**, **property:**
 Probabilistic
-  **to:**, **attachment:**, **mimetype:**
+  **body:**, **to:**, **attachment:**, **mimetype:**
 Special
    **from:**, **query:**, **subject:**
 
diff --git a/lib/database-private.h b/lib/database-private.h
index a499b259..293f2db4 100644
--- a/lib/database-private.h
+++ b/lib/database-private.h
@@ -108,6 +108,12 @@ enum _notmuch_features {
      *
      * Introduced: version 3. */
     NOTMUCH_FEATURE_LAST_MOD = 1 << 6,
+
+    /* If set, unprefixed terms are stored only for the message body,
+     * not for headers.
+     *
+     * Introduced: version 3. */
+    NOTMUCH_FEATURE_UNPREFIX_BODY_ONLY = 1 << 7,
 };
 
 /* In C++, a named enum is its own type, so define bitwise operators
diff --git a/lib/database.cc b/lib/database.cc
index 09ab9cb0..508dc94c 100644
--- a/lib/database.cc
+++ b/lib/database.cc
@@ -122,9 +122,12 @@ typedef struct {
  *	LAST_MOD:	The revision number as of the last tag or
  *			filename change.
  *
- * In addition, terms from the content of the message are added with
- * "from", "to", "attachment", and "subject" prefixes for use by the
- * user in searching.
+ * The prefixed terms described above are also searchable without an
+ * explicit field name, but as of notmuch 0.29 this is due to
+ * query-parser setup, not extra terms in the database.  In addition,
+ * terms from the content of the message are added without a prefix
+ * for use by the user in searching. Note that the prefix name "body"
+ * is used to refer to the empty prefix string in the database.
  *
  * The path of the containing folder is added with the "folder" prefix
  * (see _notmuch_message_add_folder_terms).  Sub-paths of the the path
@@ -266,6 +269,8 @@ prefix_t prefix_table[] = {
     { "directory",		"XDIRECTORY",	NOTMUCH_FIELD_NO_FLAGS },
     { "file-direntry",		"XFDIRENTRY",	NOTMUCH_FIELD_NO_FLAGS },
     { "directory-direntry",	"XDDIRENTRY",	NOTMUCH_FIELD_NO_FLAGS },
+    { "body",			"",		NOTMUCH_FIELD_EXTERNAL |
+						NOTMUCH_FIELD_PROBABILISTIC},
     { "thread",			"G",		NOTMUCH_FIELD_EXTERNAL |
 						NOTMUCH_FIELD_PROCESSOR },
     { "tag",			"K",		NOTMUCH_FIELD_EXTERNAL |
@@ -309,6 +314,8 @@ prefix_t prefix_table[] = {
 static void
 _setup_query_field_default (const prefix_t *prefix, notmuch_database_t *notmuch)
 {
+    if (prefix->prefix)
+	notmuch->query_parser->add_prefix("",prefix->prefix);
     if (prefix->flags & NOTMUCH_FIELD_PROBABILISTIC)
 	notmuch->query_parser->add_prefix (prefix->name, prefix->prefix);
     else
@@ -333,6 +340,8 @@ _setup_query_field (const prefix_t *prefix, notmuch_database_t *notmuch)
 					    *notmuch->query_parser, notmuch))->release ();
 
 	/* we treat all field-processor fields as boolean in order to get the raw input */
+	if (prefix->prefix)
+	    notmuch->query_parser->add_prefix("",prefix->prefix);
 	notmuch->query_parser->add_boolean_prefix (prefix->name, fp);
     } else {
 	_setup_query_field_default (prefix, notmuch);
@@ -390,6 +399,10 @@ static const struct {
       "indexed MIME types", "w"},
     { NOTMUCH_FEATURE_LAST_MOD,
       "modification tracking", "w"},
+    /* Existing databases will work fine for all queries not involving
+     * 'body:' */
+    { NOTMUCH_FEATURE_UNPREFIX_BODY_ONLY,
+      "index body and headers separately", "w"},
 };
 
 const char *
@@ -663,6 +676,7 @@ notmuch_database_create_verbose (const char *path,
      * new databases have them. */
     notmuch->features |= NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES;
     notmuch->features |= NOTMUCH_FEATURE_INDEXED_MIMETYPES;
+    notmuch->features |= NOTMUCH_FEATURE_UNPREFIX_BODY_ONLY;
 
     status = notmuch_database_upgrade (notmuch, NULL, NULL);
     if (status) {
diff --git a/lib/message.cc b/lib/message.cc
index 6f2f6345..38a48933 100644
--- a/lib/message.cc
+++ b/lib/message.cc
@@ -1419,8 +1419,9 @@ _notmuch_message_add_term (notmuch_message_t *message,
 }
 
 /* Parse 'text' and add a term to 'message' for each parsed word. Each
- * term will be added both prefixed (if prefix_name is not NULL) and
- * also non-prefixed). */
+ * term will be added with the appropriate prefix if prefix_name is
+ * non-NULL.
+ */
 notmuch_private_status_t
 _notmuch_message_gen_terms (notmuch_message_t *message,
 			    const char *prefix_name,
@@ -1432,22 +1433,17 @@ _notmuch_message_gen_terms (notmuch_message_t *message,
 	return NOTMUCH_PRIVATE_STATUS_NULL_POINTER;
 
     term_gen->set_document (message->doc);
+    term_gen->set_termpos (message->termpos);
 
     if (prefix_name) {
-	const char *prefix = _find_prefix (prefix_name);
-
-	term_gen->set_termpos (message->termpos);
-	term_gen->index_text (text, 1, prefix);
-	/* Create a gap between this an the next terms so they don't
-	 * appear to be a phrase. */
-	message->termpos = term_gen->get_termpos () + 100;
-
 	_notmuch_message_invalidate_metadata (message, prefix_name);
+	term_gen->index_text (text, 1, _find_prefix (prefix_name));
+    } else {
+	term_gen->index_text (text);
     }
 
-    term_gen->set_termpos (message->termpos);
-    term_gen->index_text (text);
-    /* Create a term gap, as above. */
+    /* Create a gap between this an the next terms so they don't
+     * appear to be a phrase. */
     message->termpos = term_gen->get_termpos () + 100;
 
     return NOTMUCH_PRIVATE_STATUS_SUCCESS;
diff --git a/test/T530-upgrade.sh b/test/T530-upgrade.sh
index 69ebec68..2124dde2 100755
--- a/test/T530-upgrade.sh
+++ b/test/T530-upgrade.sh
@@ -117,4 +117,20 @@ MAIL_DIR/bar/new/21:2,
 MAIL_DIR/bar/new/22:2,
 MAIL_DIR/cur/51:2,"
 
+test_begin_subtest "body: same as unprefixed before reindex"
+notmuch search --output=messages body:close > OUTPUT
+notmuch search --output=messages close  > EXPECTED
+test_expect_equal_file EXPECTED OUTPUT
+
+test_begin_subtest "body: subset of unprefixed after reindex"
+notmuch reindex '*'
+notmuch search --output=messages body:close | sort > BODY
+notmuch search --output=messages close | sort > UNPREFIXED
+diff -e UNPREFIXED BODY | cut -c2- > OUTPUT
+cat <<EOF > EXPECTED
+d
+d
+EOF
+test_expect_equal_file EXPECTED OUTPUT
+
 test_done
diff --git a/test/T730-body.sh b/test/T730-body.sh
new file mode 100755
index 00000000..548b30a4
--- /dev/null
+++ b/test/T730-body.sh
@@ -0,0 +1,43 @@
+#!/usr/bin/env bash
+test_description='search body'
+. $(dirname "$0")/test-lib.sh || exit 1
+
+add_message "[body]=thebody-1" "[subject]=subject-1"
+add_message "[body]=nothing-to-see-here-1" "[subject]=thebody-1"
+
+test_begin_subtest 'search with body: prefix'
+notmuch search body:thebody | notmuch_search_sanitize > OUTPUT
+cat <<EOF > EXPECTED
+thread:XXX   2001-01-05 [1/1] Notmuch Test Suite; subject-1 (inbox unread)
+EOF
+test_expect_equal_file EXPECTED OUTPUT
+
+test_begin_subtest 'search without body: prefix'
+notmuch search thebody | notmuch_search_sanitize > OUTPUT
+cat <<EOF > EXPECTED
+thread:XXX   2001-01-05 [1/1] Notmuch Test Suite; subject-1 (inbox unread)
+thread:XXX   2001-01-05 [1/1] Notmuch Test Suite; thebody-1 (inbox unread)
+EOF
+test_expect_equal_file EXPECTED OUTPUT
+
+test_begin_subtest 'negated body: prefix'
+notmuch search thebody and not body:thebody | notmuch_search_sanitize > OUTPUT
+cat <<EOF > EXPECTED
+thread:XXX   2001-01-05 [1/1] Notmuch Test Suite; thebody-1 (inbox unread)
+EOF
+test_expect_equal_file EXPECTED OUTPUT
+
+test_begin_subtest 'search unprefixed for prefixed term'
+notmuch search subject | notmuch_search_sanitize > OUTPUT
+cat <<EOF > EXPECTED
+thread:XXX   2001-01-05 [1/1] Notmuch Test Suite; subject-1 (inbox unread)
+EOF
+test_expect_equal_file EXPECTED OUTPUT
+
+test_begin_subtest 'search with body: prefix for term only in subject'
+notmuch search body:subject | notmuch_search_sanitize > OUTPUT
+cat <<EOF > EXPECTED
+EOF
+test_expect_equal_file EXPECTED OUTPUT
+
+test_done
-- 
2.20.1