[PATCH v4 00/11] Implement and use database "features"

[PATCH v4 00/11] Implement and use database "features"

[Austin Clements]

This is v4 of id:1406859003-11561-1-git-send-email-amdragon@mit.edu.
In addition to rebasing to current master, this makes several tidying
changes: it gives the features enum a name for better
self-documentation and type-safety; improves the robustness of
_parse_features to NULL pointers; requests upgrades if the database
version is old, even if it supports all current features; improves
various comments; and fixes some style errors.

The diff from v3 is below.  Most of this diff relates to giving the
enum a name, since it has to move above struct _notmuch_database and
we need to define bitwise operators for C++.

diff --git a/lib/database-private.h b/lib/database-private.h
index 2ffab33..ca0751c 100644
--- a/lib/database-private.h
+++ b/lib/database-private.h
@@ -36,36 +36,30 @@
 
 #pragma GCC visibility push(hidden)
 
-struct _notmuch_database {
-    notmuch_bool_t exception_reported;
-
-    char *path;
-
-    notmuch_database_mode_t mode;
-    int atomic_nesting;
-    Xapian::Database *xapian_db;
-
-    /* Bit mask of features used by this database.  Features are
-     * named, independent aspects of the database schema.  This is a
-     * bitwise-OR of NOTMUCH_FEATURE_* values (below). */
-    unsigned int features;
-
-    unsigned int last_doc_id;
-    uint64_t last_thread_id;
-
-    Xapian::QueryParser *query_parser;
-    Xapian::TermGenerator *term_gen;
-    Xapian::ValueRangeProcessor *value_range_processor;
-    Xapian::ValueRangeProcessor *date_range_processor;
-};
-
-/* Bit masks for _notmuch_database::features. */
-enum {
+/* Bit masks for _notmuch_database::features.  Features are named,
+ * independent aspects of the database schema.
+ *
+ * A database stores the set of features that it "uses" (implicitly
+ * before database version 3 and explicitly as of version 3).
+ *
+ * A given library version will "recognize" a particular set of
+ * features; if a database uses a feature that the library does not
+ * recognize, the library will refuse to open it.  It is assumed the
+ * set of recognized features grows monotonically over time.  A
+ * library version will "implement" some subset of the recognized
+ * features: some operations may require that the database use (or not
+ * use) some feature, while other operations may support both
+ * databases that use and that don't use some feature.
+ *
+ * On disk, the database stores string names for these features (see
+ * the feature_names array).  These enum bit values are never
+ * persisted to disk and may change freely.
+ */
+enum _notmuch_features {
     /* If set, file names are stored in "file-direntry" terms.  If
      * unset, file names are stored in document data.
      *
-     * Introduced: version 1.  Implementation support: both for read;
-     * required for write. */
+     * Introduced: version 1. */
     NOTMUCH_FEATURE_FILE_TERMS = 1 << 0,
 
     /* If set, directory timestamps are stored in documents with
@@ -73,7 +67,7 @@ enum {
      * timestamps are stored in documents with XTIMESTAMP terms and
      * absolute paths.
      *
-     * Introduced: version 1.  Implementation support: required. */
+     * Introduced: version 1. */
     NOTMUCH_FEATURE_DIRECTORY_DOCS = 1 << 1,
 
     /* If set, the from, subject, and message-id headers are stored in
@@ -82,7 +76,6 @@ enum {
      * retrieved from the message file.
      *
      * Introduced: optional in version 1, required as of version 3.
-     * Implementation support: both.
      */
     NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES = 1 << 2,
 
@@ -90,13 +83,71 @@ enum {
      * unset, folder terms are probabilistic and stemmed and path
      * terms do not exist.
      *
-     * Introduced: version 2.  Implementation support: required. */
+     * Introduced: version 2. */
     NOTMUCH_FEATURE_BOOL_FOLDER = 1 << 3,
 };
 
+/* In C++, a named enum is its own type, so define bitwise operators
+ * on _notmuch_features. */
+inline _notmuch_features
+operator|(_notmuch_features a, _notmuch_features b)
+{
+    return static_cast<_notmuch_features>(
+	static_cast<unsigned>(a) | static_cast<unsigned>(b));
+}
+
+inline _notmuch_features
+operator&(_notmuch_features a, _notmuch_features b)
+{
+    return static_cast<_notmuch_features>(
+	static_cast<unsigned>(a) & static_cast<unsigned>(b));
+}
+
+inline _notmuch_features
+operator~(_notmuch_features a)
+{
+    return static_cast<_notmuch_features>(~static_cast<unsigned>(a));
+}
+
+inline _notmuch_features&
+operator|=(_notmuch_features &a, _notmuch_features b)
+{
+    a = a | b;
+    return a;
+}
+
+inline _notmuch_features&
+operator&=(_notmuch_features &a, _notmuch_features b)
+{
+    a = a & b;
+    return a;
+}
+
+struct _notmuch_database {
+    notmuch_bool_t exception_reported;
+
+    char *path;
+
+    notmuch_database_mode_t mode;
+    int atomic_nesting;
+    Xapian::Database *xapian_db;
+
+    /* Bit mask of features used by this database.  This is a
+     * bitwise-OR of NOTMUCH_FEATURE_* values (above). */
+    enum _notmuch_features features;
+
+    unsigned int last_doc_id;
+    uint64_t last_thread_id;
+
+    Xapian::QueryParser *query_parser;
+    Xapian::TermGenerator *term_gen;
+    Xapian::ValueRangeProcessor *value_range_processor;
+    Xapian::ValueRangeProcessor *date_range_processor;
+};
+
 /* Prior to database version 3, features were implied by the database
  * version number, so hard-code them for earlier versions. */
-#define NOTMUCH_FEATURES_V0 (0)
+#define NOTMUCH_FEATURES_V0 ((enum _notmuch_features)0)
 #define NOTMUCH_FEATURES_V1 (NOTMUCH_FEATURES_V0 | NOTMUCH_FEATURE_FILE_TERMS | \
 			     NOTMUCH_FEATURE_DIRECTORY_DOCS)
 #define NOTMUCH_FEATURES_V2 (NOTMUCH_FEATURES_V1 | NOTMUCH_FEATURE_BOOL_FOLDER)
diff --git a/lib/database.cc b/lib/database.cc
index 63257c2..5116188 100644
--- a/lib/database.cc
+++ b/lib/database.cc
@@ -266,10 +266,9 @@ _find_prefix (const char *name)
     return "";
 }
 
-static const struct
-{
+static const struct {
     /* NOTMUCH_FEATURE_* value. */
-    unsigned int value;
+    _notmuch_features value;
     /* Feature name as it appears in the database.  This name should
      * be appropriate for displaying to the user if an older version
      * of notmuch doesn't support this feature. */
@@ -277,12 +276,16 @@ static const struct
     /* Compatibility flags when this feature is declared. */
     const char *flags;
 } feature_names[] = {
-    {NOTMUCH_FEATURE_FILE_TERMS,             "multiple paths per message", "rw"},
-    {NOTMUCH_FEATURE_DIRECTORY_DOCS,         "relative directory paths", "rw"},
+    { NOTMUCH_FEATURE_FILE_TERMS,
+      "multiple paths per message", "rw" },
+    { NOTMUCH_FEATURE_DIRECTORY_DOCS,
+      "relative directory paths", "rw" },
     /* Header values are not required for reading a database because a
      * reader can just refer to the message file. */
-    {NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES, "from/subject/message-ID in database", "w"},
-    {NOTMUCH_FEATURE_BOOL_FOLDER,            "exact folder:/path: search", "rw"},
+    { NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES,
+      "from/subject/message-ID in database", "w" },
+    { NOTMUCH_FEATURE_BOOL_FOLDER,
+      "exact folder:/path: search", "rw" },
 };
 
 const char *
@@ -658,6 +661,7 @@ _notmuch_database_ensure_writable (notmuch_database_t *notmuch)
 }
 
 /* Parse a database features string from the given database version.
+ * Returns the feature bit set.
  *
  * For version < 3, this ignores the features string and returns a
  * hard-coded set of features.
@@ -665,13 +669,15 @@ _notmuch_database_ensure_writable (notmuch_database_t *notmuch)
  * If there are unrecognized features that are required to open the
  * database in mode (which should be 'r' or 'w'), return a
  * comma-separated list of unrecognized but required features in
- * *incompat_out, which will be allocated from ctx.
+ * *incompat_out suitable for presenting to the user.  *incompat_out
+ * will be allocated from ctx.
  */
-static unsigned int
+static _notmuch_features
 _parse_features (const void *ctx, const char *features, unsigned int version,
 		 char mode, char **incompat_out)
 {
-    unsigned int res = 0, namelen, i;
+    _notmuch_features res = static_cast<_notmuch_features>(0);
+    unsigned int namelen, i;
     size_t llen = 0;
     const char *flags;
 
@@ -699,7 +705,7 @@ _parse_features (const void *ctx, const char *features, unsigned int version,
 	    }
 	}
 
-	if (i == ARRAY_SIZE (feature_names)) {
+	if (i == ARRAY_SIZE (feature_names) && incompat_out) {
 	    /* Unrecognized feature */
 	    const char *have = strchr (flags, mode);
 	    if (have && have < features + llen) {
@@ -1167,7 +1173,8 @@ notmuch_bool_t
 notmuch_database_needs_upgrade (notmuch_database_t *notmuch)
 {
     return notmuch->mode == NOTMUCH_DATABASE_MODE_READ_WRITE &&
-	(NOTMUCH_FEATURES_CURRENT & ~notmuch->features);
+	((NOTMUCH_FEATURES_CURRENT & ~notmuch->features) ||
+	 (notmuch_database_get_version (notmuch) < NOTMUCH_DATABASE_VERSION));
 }
 
 static volatile sig_atomic_t do_progress_notify = 0;
@@ -1202,7 +1209,7 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
     struct sigaction action;
     struct itimerval timerval;
     notmuch_bool_t timer_is_active = FALSE;
-    unsigned int target_features, new_features;
+    enum _notmuch_features target_features, new_features;
     notmuch_status_t status;
     unsigned int count = 0, total = 0;
 
diff --git a/lib/notmuch.h b/lib/notmuch.h
index d771eb2..21a5225 100644
--- a/lib/notmuch.h
+++ b/lib/notmuch.h
@@ -352,12 +352,14 @@ unsigned int
 notmuch_database_get_version (notmuch_database_t *database);
 
 /**
- * Is the database behind the latest supported database version?
+ * Can the database be upgraded to a newer database version?
  *
  * If this function returns TRUE, then the caller may call
  * notmuch_database_upgrade to upgrade the database.  If the caller
  * does not upgrade an out-of-date database, then some functions may
- * fail with NOTMUCH_STATUS_UPGRADE_REQUIRED.
+ * fail with NOTMUCH_STATUS_UPGRADE_REQUIRED.  This always returns
+ * FALSE for a read-only database because there's no way to upgrade a
+ * read-only database.
  */
 notmuch_bool_t
 notmuch_database_needs_upgrade (notmuch_database_t *database);

[PATCH v4 01/11] new: Don't report version after upgrade

[Austin Clements]

The version number has always been pretty meaningless to the user and
it's about to become even more meaningless with the introduction of
"features".  Hopefully, the database will remain on version 3 for some
time to come; however, the introduction of new features over time in
version 3 will necessitate upgrades within version 3.  It would be
confusing if we always tell the user they've been "upgraded to version
3".  If the user wants to know what's new, they should read the news.
---
 notmuch-new.c        | 3 +--
 test/T530-upgrade.sh | 2 +-
 2 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/notmuch-new.c b/notmuch-new.c
index 5691005..ddf42c1 100644
--- a/notmuch-new.c
+++ b/notmuch-new.c
@@ -1031,8 +1031,7 @@ notmuch_new_command (notmuch_config_t *config, int argc, char *argv[])
 		return EXIT_FAILURE;
 	    }
 	    if (add_files_state.verbosity >= VERBOSITY_NORMAL)
-		printf ("Your notmuch database has now been upgraded to database format version %u.\n",
-		    notmuch_database_get_version (notmuch));
+		printf ("Your notmuch database has now been upgraded.\n");
 	}
 
 	add_files_state.total_files = 0;
diff --git a/test/T530-upgrade.sh b/test/T530-upgrade.sh
index 7d5d5aa..c4c4ac8 100755
--- a/test/T530-upgrade.sh
+++ b/test/T530-upgrade.sh
@@ -33,7 +33,7 @@ test_expect_equal "$output" "\
 Welcome to a new version of notmuch! Your database will now be upgraded.
 This process is safe to interrupt.
 Backing up tags to FILENAME
-Your notmuch database has now been upgraded to database format version 2.
+Your notmuch database has now been upgraded.
 No new mail."
 
 test_begin_subtest "tag backup matches pre-upgrade dump"
-- 
2.0.0

[PATCH v4 02/11] lib: Database version 3: Introduce fine-grained "features"

[Austin Clements]

Previously, our database schema was versioned by a single number.
Each database schema change had to occur "atomically" in Notmuch's
development history: before some commit, Notmuch used version N, after
that commit, it used version N+1.  Hence, each new schema version
could introduce only one change, the task of developing a schema
change fell on a single person, and it all had to happen and be
perfect in a single commit series.  This made introducing a new schema
version hard.  We've seen only two schema changes in the history of
Notmuch.

This commit introduces database schema version 3; hopefully the last
schema version we'll need for a while.  With this version, we switch
from a single version number to "features": a set of named,
independent aspects of the database schema.

Features should make backwards compatibility easier.  For many things,
it should be easy to support databases both with and without a
feature, which will allow us to make upgrades optional and will enable
"unstable" features that can be developed and tested over time.

Features also make forwards compatibility easier.  The features
recorded in a database include "compatibility flags," which can
indicate to an older version of Notmuch when it must support a given
feature to open the database for read or for write.  This lets us
replace the old vague "I don't recognize this version, so something
might go wrong, but I promise to try my best" warnings upon opening a
database with an unknown version with precise errors.  If a database
is safe to open for read/write despite unknown features, an older
version will know that and issue no message at all.  If the database
is not safe to open for read/write because of unknown features, an
older version will know that, too, and can tell the user exactly which
required features it lacks support for.
---
 lib/database-private.h | 108 ++++++++++++++++++++++++++-
 lib/database.cc        | 197 ++++++++++++++++++++++++++++++++++++++++---------
 2 files changed, 271 insertions(+), 34 deletions(-)

diff --git a/lib/database-private.h b/lib/database-private.h
index d3e65fd..ca0751c 100644
--- a/lib/database-private.h
+++ b/lib/database-private.h
@@ -36,16 +36,106 @@
 
 #pragma GCC visibility push(hidden)
 
+/* Bit masks for _notmuch_database::features.  Features are named,
+ * independent aspects of the database schema.
+ *
+ * A database stores the set of features that it "uses" (implicitly
+ * before database version 3 and explicitly as of version 3).
+ *
+ * A given library version will "recognize" a particular set of
+ * features; if a database uses a feature that the library does not
+ * recognize, the library will refuse to open it.  It is assumed the
+ * set of recognized features grows monotonically over time.  A
+ * library version will "implement" some subset of the recognized
+ * features: some operations may require that the database use (or not
+ * use) some feature, while other operations may support both
+ * databases that use and that don't use some feature.
+ *
+ * On disk, the database stores string names for these features (see
+ * the feature_names array).  These enum bit values are never
+ * persisted to disk and may change freely.
+ */
+enum _notmuch_features {
+    /* If set, file names are stored in "file-direntry" terms.  If
+     * unset, file names are stored in document data.
+     *
+     * Introduced: version 1. */
+    NOTMUCH_FEATURE_FILE_TERMS = 1 << 0,
+
+    /* If set, directory timestamps are stored in documents with
+     * XDIRECTORY terms and relative paths.  If unset, directory
+     * timestamps are stored in documents with XTIMESTAMP terms and
+     * absolute paths.
+     *
+     * Introduced: version 1. */
+    NOTMUCH_FEATURE_DIRECTORY_DOCS = 1 << 1,
+
+    /* If set, the from, subject, and message-id headers are stored in
+     * message document values.  If unset, message documents *may*
+     * have these values, but if the value is empty, it must be
+     * retrieved from the message file.
+     *
+     * Introduced: optional in version 1, required as of version 3.
+     */
+    NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES = 1 << 2,
+
+    /* If set, folder terms are boolean and path terms exist.  If
+     * unset, folder terms are probabilistic and stemmed and path
+     * terms do not exist.
+     *
+     * Introduced: version 2. */
+    NOTMUCH_FEATURE_BOOL_FOLDER = 1 << 3,
+};
+
+/* In C++, a named enum is its own type, so define bitwise operators
+ * on _notmuch_features. */
+inline _notmuch_features
+operator|(_notmuch_features a, _notmuch_features b)
+{
+    return static_cast<_notmuch_features>(
+	static_cast<unsigned>(a) | static_cast<unsigned>(b));
+}
+
+inline _notmuch_features
+operator&(_notmuch_features a, _notmuch_features b)
+{
+    return static_cast<_notmuch_features>(
+	static_cast<unsigned>(a) & static_cast<unsigned>(b));
+}
+
+inline _notmuch_features
+operator~(_notmuch_features a)
+{
+    return static_cast<_notmuch_features>(~static_cast<unsigned>(a));
+}
+
+inline _notmuch_features&
+operator|=(_notmuch_features &a, _notmuch_features b)
+{
+    a = a | b;
+    return a;
+}
+
+inline _notmuch_features&
+operator&=(_notmuch_features &a, _notmuch_features b)
+{
+    a = a & b;
+    return a;
+}
+
 struct _notmuch_database {
     notmuch_bool_t exception_reported;
 
     char *path;
 
-    notmuch_bool_t needs_upgrade;
     notmuch_database_mode_t mode;
     int atomic_nesting;
     Xapian::Database *xapian_db;
 
+    /* Bit mask of features used by this database.  This is a
+     * bitwise-OR of NOTMUCH_FEATURE_* values (above). */
+    enum _notmuch_features features;
+
     unsigned int last_doc_id;
     uint64_t last_thread_id;
 
@@ -55,6 +145,22 @@ struct _notmuch_database {
     Xapian::ValueRangeProcessor *date_range_processor;
 };
 
+/* Prior to database version 3, features were implied by the database
+ * version number, so hard-code them for earlier versions. */
+#define NOTMUCH_FEATURES_V0 ((enum _notmuch_features)0)
+#define NOTMUCH_FEATURES_V1 (NOTMUCH_FEATURES_V0 | NOTMUCH_FEATURE_FILE_TERMS | \
+			     NOTMUCH_FEATURE_DIRECTORY_DOCS)
+#define NOTMUCH_FEATURES_V2 (NOTMUCH_FEATURES_V1 | NOTMUCH_FEATURE_BOOL_FOLDER)
+
+/* Current database features.  If any of these are missing from a
+ * database, request an upgrade.
+ * NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES is not included because
+ * upgrade doesn't currently introduce the feature (though brand new
+ * databases will have it). */
+#define NOTMUCH_FEATURES_CURRENT \
+    (NOTMUCH_FEATURE_FILE_TERMS | NOTMUCH_FEATURE_DIRECTORY_DOCS | \
+     NOTMUCH_FEATURE_BOOL_FOLDER)
+
 /* Return the list of terms from the given iterator matching a prefix.
  * The prefix will be stripped from the strings in the returned list.
  * The list will be allocated using ctx as the talloc context.
diff --git a/lib/database.cc b/lib/database.cc
index 9c0952a..2b566f7 100644
--- a/lib/database.cc
+++ b/lib/database.cc
@@ -20,6 +20,7 @@
 
 #include "database-private.h"
 #include "parse-time-vrp.h"
+#include "string-util.h"
 
 #include <iostream>
 
@@ -42,7 +43,7 @@ typedef struct {
     const char *prefix;
 } prefix_t;
 
-#define NOTMUCH_DATABASE_VERSION 2
+#define NOTMUCH_DATABASE_VERSION 3
 
 #define STRINGIFY(s) _SUB_STRINGIFY(s)
 #define _SUB_STRINGIFY(s) #s
@@ -154,6 +155,17 @@ typedef struct {
  *			changes are made to the database (such as by
  *			indexing new fields).
  *
+ *	features	The set of features supported by this
+ *			database. This consists of a set of
+ *			'\n'-separated lines, where each is a feature
+ *			name, a '\t', and compatibility flags.  If the
+ *			compatibility flags contain 'w', then the
+ *			opener must support this feature to safely
+ *			write this database.  If the compatibility
+ *			flags contain 'r', then the opener must
+ *			support this feature to read this database.
+ *			Introduced in database version 3.
+ *
  *	last_thread_id	The last thread ID generated. This is stored
  *			as a 16-byte hexadecimal ASCII representation
  *			of a 64-bit unsigned integer. The first ID
@@ -254,6 +266,28 @@ _find_prefix (const char *name)
     return "";
 }
 
+static const struct {
+    /* NOTMUCH_FEATURE_* value. */
+    _notmuch_features value;
+    /* Feature name as it appears in the database.  This name should
+     * be appropriate for displaying to the user if an older version
+     * of notmuch doesn't support this feature. */
+    const char *name;
+    /* Compatibility flags when this feature is declared. */
+    const char *flags;
+} feature_names[] = {
+    { NOTMUCH_FEATURE_FILE_TERMS,
+      "multiple paths per message", "rw" },
+    { NOTMUCH_FEATURE_DIRECTORY_DOCS,
+      "relative directory paths", "rw" },
+    /* Header values are not required for reading a database because a
+     * reader can just refer to the message file. */
+    { NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES,
+      "from/subject/message-ID in database", "w" },
+    { NOTMUCH_FEATURE_BOOL_FOLDER,
+      "exact folder:/path: search", "rw" },
+};
+
 const char *
 notmuch_status_to_string (notmuch_status_t status)
 {
@@ -591,6 +625,11 @@ notmuch_database_create (const char *path, notmuch_database_t **database)
 				    &notmuch);
     if (status)
 	goto DONE;
+
+    /* Upgrade doesn't add this feature to existing databases, but new
+     * databases have it. */
+    notmuch->features |= NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES;
+
     status = notmuch_database_upgrade (notmuch, NULL, NULL);
     if (status) {
 	notmuch_database_close(notmuch);
@@ -619,6 +658,83 @@ _notmuch_database_ensure_writable (notmuch_database_t *notmuch)
     return NOTMUCH_STATUS_SUCCESS;
 }
 
+/* Parse a database features string from the given database version.
+ * Returns the feature bit set.
+ *
+ * For version < 3, this ignores the features string and returns a
+ * hard-coded set of features.
+ *
+ * If there are unrecognized features that are required to open the
+ * database in mode (which should be 'r' or 'w'), return a
+ * comma-separated list of unrecognized but required features in
+ * *incompat_out suitable for presenting to the user.  *incompat_out
+ * will be allocated from ctx.
+ */
+static _notmuch_features
+_parse_features (const void *ctx, const char *features, unsigned int version,
+		 char mode, char **incompat_out)
+{
+    _notmuch_features res = static_cast<_notmuch_features>(0);
+    unsigned int namelen, i;
+    size_t llen = 0;
+    const char *flags;
+
+    /* Prior to database version 3, features were implied by the
+     * version number. */
+    if (version == 0)
+	return NOTMUCH_FEATURES_V0;
+    else if (version == 1)
+	return NOTMUCH_FEATURES_V1;
+    else if (version == 2)
+	return NOTMUCH_FEATURES_V2;
+
+    /* Parse the features string */
+    while ((features = strtok_len_c (features + llen, "\n", &llen)) != NULL) {
+	flags = strchr (features, '\t');
+	if (! flags || flags > features + llen)
+	    continue;
+	namelen = flags - features;
+
+	for (i = 0; i < ARRAY_SIZE (feature_names); ++i) {
+	    if (strlen (feature_names[i].name) == namelen &&
+		strncmp (feature_names[i].name, features, namelen) == 0) {
+		res |= feature_names[i].value;
+		break;
+	    }
+	}
+
+	if (i == ARRAY_SIZE (feature_names) && incompat_out) {
+	    /* Unrecognized feature */
+	    const char *have = strchr (flags, mode);
+	    if (have && have < features + llen) {
+		/* This feature is required to access this database in
+		 * 'mode', but we don't understand it. */
+		if (! *incompat_out)
+		    *incompat_out = talloc_strdup (ctx, "");
+		*incompat_out = talloc_asprintf_append_buffer (
+		    *incompat_out, "%s%.*s", **incompat_out ? ", " : "",
+		    namelen, features);
+	    }
+	}
+    }
+
+    return res;
+}
+
+static char *
+_print_features (const void *ctx, unsigned int features)
+{
+    unsigned int i;
+    char *res = talloc_strdup (ctx, "");
+
+    for (i = 0; i < ARRAY_SIZE (feature_names); ++i)
+	if (features & feature_names[i].value)
+	    res = talloc_asprintf_append_buffer (
+		res, "%s\t%s\n", feature_names[i].name, feature_names[i].flags);
+
+    return res;
+}
+
 notmuch_status_t
 notmuch_database_open (const char *path,
 		       notmuch_database_mode_t mode,
@@ -627,7 +743,7 @@ notmuch_database_open (const char *path,
     notmuch_status_t status = NOTMUCH_STATUS_SUCCESS;
     void *local = talloc_new (NULL);
     notmuch_database_t *notmuch = NULL;
-    char *notmuch_path, *xapian_path;
+    char *notmuch_path, *xapian_path, *incompat_features;
     struct stat st;
     int err;
     unsigned int i, version;
@@ -677,7 +793,6 @@ notmuch_database_open (const char *path,
     if (notmuch->path[strlen (notmuch->path) - 1] == '/')
 	notmuch->path[strlen (notmuch->path) - 1] = '\0';
 
-    notmuch->needs_upgrade = FALSE;
     notmuch->mode = mode;
     notmuch->atomic_nesting = 0;
     try {
@@ -686,37 +801,44 @@ notmuch_database_open (const char *path,
 	if (mode == NOTMUCH_DATABASE_MODE_READ_WRITE) {
 	    notmuch->xapian_db = new Xapian::WritableDatabase (xapian_path,
 							       Xapian::DB_CREATE_OR_OPEN);
-	    version = notmuch_database_get_version (notmuch);
-
-	    if (version > NOTMUCH_DATABASE_VERSION) {
-		fprintf (stderr,
-			 "Error: Notmuch database at %s\n"
-			 "       has a newer database format version (%u) than supported by this\n"
-			 "       version of notmuch (%u). Refusing to open this database in\n"
-			 "       read-write mode.\n",
-			 notmuch_path, version, NOTMUCH_DATABASE_VERSION);
-		notmuch->mode = NOTMUCH_DATABASE_MODE_READ_ONLY;
-		notmuch_database_destroy (notmuch);
-		notmuch = NULL;
-		status = NOTMUCH_STATUS_FILE_ERROR;
-		goto DONE;
-	    }
-
-	    if (version < NOTMUCH_DATABASE_VERSION)
-		notmuch->needs_upgrade = TRUE;
 	} else {
 	    notmuch->xapian_db = new Xapian::Database (xapian_path);
-	    version = notmuch_database_get_version (notmuch);
-	    if (version > NOTMUCH_DATABASE_VERSION)
-	    {
-		fprintf (stderr,
-			 "Warning: Notmuch database at %s\n"
-			 "         has a newer database format version (%u) than supported by this\n"
-			 "         version of notmuch (%u). Some operations may behave incorrectly,\n"
-			 "         (but the database will not be harmed since it is being opened\n"
-			 "         in read-only mode).\n",
-			 notmuch_path, version, NOTMUCH_DATABASE_VERSION);
-	    }
+	}
+
+	/* Check version.  As of database version 3, we represent
+	 * changes in terms of features, so assume a version bump
+	 * means a dramatically incompatible change. */
+	version = notmuch_database_get_version (notmuch);
+	if (version > NOTMUCH_DATABASE_VERSION) {
+	    fprintf (stderr,
+		     "Error: Notmuch database at %s\n"
+		     "       has a newer database format version (%u) than supported by this\n"
+		     "       version of notmuch (%u).\n",
+		     notmuch_path, version, NOTMUCH_DATABASE_VERSION);
+	    notmuch->mode = NOTMUCH_DATABASE_MODE_READ_ONLY;
+	    notmuch_database_destroy (notmuch);
+	    notmuch = NULL;
+	    status = NOTMUCH_STATUS_FILE_ERROR;
+	    goto DONE;
+	}
+
+	/* Check features. */
+	incompat_features = NULL;
+	notmuch->features = _parse_features (
+	    local, notmuch->xapian_db->get_metadata ("features").c_str (),
+	    version, mode == NOTMUCH_DATABASE_MODE_READ_WRITE ? 'w' : 'r',
+	    &incompat_features);
+	if (incompat_features) {
+	    fprintf (stderr,
+		     "Error: Notmuch database at %s\n"
+		     "       requires features (%s)\n"
+		     "       not supported by this version of notmuch.\n",
+		     notmuch_path, incompat_features);
+	    notmuch->mode = NOTMUCH_DATABASE_MODE_READ_ONLY;
+	    notmuch_database_destroy (notmuch);
+	    notmuch = NULL;
+	    status = NOTMUCH_STATUS_FILE_ERROR;
+	    goto DONE;
 	}
 
 	notmuch->last_doc_id = notmuch->xapian_db->get_lastdocid ();
@@ -1048,7 +1170,9 @@ notmuch_database_get_version (notmuch_database_t *notmuch)
 notmuch_bool_t
 notmuch_database_needs_upgrade (notmuch_database_t *notmuch)
 {
-    return notmuch->needs_upgrade;
+    return notmuch->mode == NOTMUCH_DATABASE_MODE_READ_WRITE &&
+	((NOTMUCH_FEATURES_CURRENT & ~notmuch->features) ||
+	 (notmuch_database_get_version (notmuch) < NOTMUCH_DATABASE_VERSION));
 }
 
 static volatile sig_atomic_t do_progress_notify = 0;
@@ -1077,6 +1201,7 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 						   double progress),
 			  void *closure)
 {
+    void *local = talloc_new (NULL);
     Xapian::WritableDatabase *db;
     struct sigaction action;
     struct itimerval timerval;
@@ -1114,6 +1239,10 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 	timer_is_active = TRUE;
     }
 
+    /* Set the target features so we write out changes in the desired
+     * format. */
+    notmuch->features |= NOTMUCH_FEATURES_CURRENT;
+
     /* Before version 1, each message document had its filename in the
      * data field. Copy that into the new format by calling
      * notmuch_message_add_filename.
@@ -1226,6 +1355,7 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 	notmuch_query_destroy (query);
     }
 
+    db->set_metadata ("features", _print_features (local, notmuch->features));
     db->set_metadata ("version", STRINGIFY (NOTMUCH_DATABASE_VERSION));
     db->flush ();
 
@@ -1302,6 +1432,7 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 	sigaction (SIGALRM, &action, NULL);
     }
 
+    talloc_free (local);
     return NOTMUCH_STATUS_SUCCESS;
 }
 
-- 
2.0.0

[PATCH v4 03/11] test: Tool to build DB with specific version and features

[Austin Clements]

This will let us test basic version and feature handling.
---
 test/.gitignore         |  1 +
 test/Makefile.local     |  4 ++++
 test/make-db-version.cc | 35 +++++++++++++++++++++++++++++++++++
 3 files changed, 40 insertions(+)
 create mode 100644 test/make-db-version.cc

diff --git a/test/.gitignore b/test/.gitignore
index b3b706d..0f7d5bf 100644
--- a/test/.gitignore
+++ b/test/.gitignore
@@ -5,5 +5,6 @@ parse-time
 random-corpus
 smtp-dummy
 symbol-test
+make-db-version
 test-results
 tmp.*
diff --git a/test/Makefile.local b/test/Makefile.local
index 916dd0b..a2d58fc 100644
--- a/test/Makefile.local
+++ b/test/Makefile.local
@@ -35,6 +35,9 @@ $(dir)/symbol-test: $(dir)/symbol-test.o lib/$(LINKER_NAME)
 $(dir)/parse-time: $(dir)/parse-time.o parse-time-string/parse-time-string.o
 	$(call quiet,CC) $^ -o $@
 
+$(dir)/make-db-version: $(dir)/make-db-version.o
+	$(call quiet,CXX) $^ -o $@ $(XAPIAN_LDFLAGS)
+
 .PHONY: test check
 
 test_main_srcs=$(dir)/arg-test.c \
@@ -43,6 +46,7 @@ test_main_srcs=$(dir)/arg-test.c \
 	      $(dir)/parse-time.c \
 	      $(dir)/smtp-dummy.c \
 	      $(dir)/symbol-test.cc \
+	      $(dir)/make-db-version.cc \
 
 test_srcs=$(test_main_srcs) $(dir)/database-test.c
 
diff --git a/test/make-db-version.cc b/test/make-db-version.cc
new file mode 100644
index 0000000..fa80cac
--- /dev/null
+++ b/test/make-db-version.cc
@@ -0,0 +1,35 @@
+/* Create an empty notmuch database with a specific version and
+ * features. */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+
+#include <xapian.h>
+
+int main(int argc, char **argv)
+{
+    if (argc != 4) {
+	fprintf (stderr, "Usage: %s mailpath version features\n", argv[0]);
+	exit (2);
+    }
+
+    std::string nmpath (argv[1]);
+    nmpath += "/.notmuch";
+    if (mkdir (nmpath.c_str (), 0777) < 0) {
+	perror (("failed to create " + nmpath).c_str ());
+	exit (1);
+    }
+
+    try {
+	Xapian::WritableDatabase db (
+	    nmpath + "/xapian", Xapian::DB_CREATE_OR_OPEN);
+	db.set_metadata ("version", argv[2]);
+	db.set_metadata ("features", argv[3]);
+	db.commit ();
+    } catch (const Xapian::Error &e) {
+	fprintf (stderr, "%s\n", e.get_description ().c_str ());
+	exit (1);
+    }
+}
-- 
2.0.0

[PATCH v4 04/11] test: Tests for future version and unknown feature handling

[Austin Clements]

---
 test/T550-db-features.sh | 48 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 48 insertions(+)
 create mode 100755 test/T550-db-features.sh

diff --git a/test/T550-db-features.sh b/test/T550-db-features.sh
new file mode 100755
index 0000000..5569768
--- /dev/null
+++ b/test/T550-db-features.sh
@@ -0,0 +1,48 @@
+#!/usr/bin/env bash
+test_description="database version and feature compatibility"
+
+. ./test-lib.sh
+
+test_begin_subtest "future database versions abort open"
+${TEST_DIRECTORY}/make-db-version ${MAIL_DIR} 9999 ""
+output=$(notmuch search x 2>&1 | sed 's/\(database at\) .*/\1 FILENAME/')
+rm -rf ${MAIL_DIR}/.notmuch
+test_expect_equal "$output" "\
+Error: Notmuch database at FILENAME
+       has a newer database format version (9999) than supported by this
+       version of notmuch (3)."
+
+test_begin_subtest "unknown 'rw' feature aborts read/write open"
+${TEST_DIRECTORY}/make-db-version ${MAIL_DIR} 3 $'test feature\trw'
+output=$(notmuch new 2>&1 | sed 's/\(database at\) .*/\1 FILENAME/')
+rm -rf ${MAIL_DIR}/.notmuch
+test_expect_equal "$output" "\
+Error: Notmuch database at FILENAME
+       requires features (test feature)
+       not supported by this version of notmuch."
+
+test_begin_subtest "unknown 'rw' feature aborts read-only open"
+${TEST_DIRECTORY}/make-db-version ${MAIL_DIR} 3 $'test feature\trw'
+output=$(notmuch search x 2>&1 | sed 's/\(database at\) .*/\1 FILENAME/')
+rm -rf ${MAIL_DIR}/.notmuch
+test_expect_equal "$output" "\
+Error: Notmuch database at FILENAME
+       requires features (test feature)
+       not supported by this version of notmuch."
+
+test_begin_subtest "unknown 'w' feature aborts read/write open"
+${TEST_DIRECTORY}/make-db-version ${MAIL_DIR} 3 $'test feature\tw'
+output=$(notmuch new 2>&1 | sed 's/\(database at\) .*/\1 FILENAME/')
+rm -rf ${MAIL_DIR}/.notmuch
+test_expect_equal "$output" "\
+Error: Notmuch database at FILENAME
+       requires features (test feature)
+       not supported by this version of notmuch."
+
+test_begin_subtest "unknown 'w' feature does not abort read-only open"
+${TEST_DIRECTORY}/make-db-version ${MAIL_DIR} 3 $'test feature\tw'
+output=$(notmuch search x 2>&1 | sed 's/\(database at\) .*/\1 FILENAME/')
+rm -rf ${MAIL_DIR}/.notmuch
+test_expect_equal "$output" ""
+
+test_done
-- 
2.0.0

[PATCH v4 05/11] lib: Simplify upgrade code using a transaction

[Austin Clements]

Previously, the upgrade was organized as two passes -- an upgrade
pass, and a separate cleanup pass -- so the database was always in a
valid state.  This change substantially simplifies this code by
performing the upgrade in a transaction and combining both passes in
to one.  This 1) eliminates a lot of duplicate code between the
passes, 2) speeds up the upgrade process, 3) makes progress reporting
more accurate, 4) eliminates the potential for stale data if the
upgrade is interrupted during the cleanup pass, and 5) makes it easier
to reason about the safety of the upgrade code.
---
 lib/database.cc | 67 ++++++---------------------------------------------------
 1 file changed, 7 insertions(+), 60 deletions(-)

diff --git a/lib/database.cc b/lib/database.cc
index 2b566f7..1e46fc8 100644
--- a/lib/database.cc
+++ b/lib/database.cc
@@ -1239,6 +1239,9 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 	timer_is_active = TRUE;
     }
 
+    /* Perform the upgrade in a transaction. */
+    db->begin_transaction (true);
+
     /* Set the target features so we write out changes in the desired
      * format. */
     notmuch->features |= NOTMUCH_FEATURES_CURRENT;
@@ -1270,6 +1273,7 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 	    filename = _notmuch_message_talloc_copy_data (message);
 	    if (filename && *filename != '\0') {
 		_notmuch_message_add_filename (message, filename);
+		_notmuch_message_clear_data (message);
 		_notmuch_message_sync (message);
 	    }
 	    talloc_free (filename);
@@ -1317,6 +1321,8 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 						       NOTMUCH_FIND_CREATE, &status);
 		notmuch_directory_set_mtime (directory, mtime);
 		notmuch_directory_destroy (directory);
+
+		db->delete_document (*p);
 	    }
 	}
     }
@@ -1357,67 +1363,8 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 
     db->set_metadata ("features", _print_features (local, notmuch->features));
     db->set_metadata ("version", STRINGIFY (NOTMUCH_DATABASE_VERSION));
-    db->flush ();
-
-    /* Now that the upgrade is complete we can remove the old data
-     * and documents that are no longer needed. */
-    if (version < 1) {
-	notmuch_query_t *query = notmuch_query_create (notmuch, "");
-	notmuch_messages_t *messages;
-	notmuch_message_t *message;
-	char *filename;
-
-	for (messages = notmuch_query_search_messages (query);
-	     notmuch_messages_valid (messages);
-	     notmuch_messages_move_to_next (messages))
-	{
-	    if (do_progress_notify) {
-		progress_notify (closure, (double) count / total);
-		do_progress_notify = 0;
-	    }
-
-	    message = notmuch_messages_get (messages);
-
-	    filename = _notmuch_message_talloc_copy_data (message);
-	    if (filename && *filename != '\0') {
-		_notmuch_message_clear_data (message);
-		_notmuch_message_sync (message);
-	    }
-	    talloc_free (filename);
-
-	    notmuch_message_destroy (message);
-	}
 
-	notmuch_query_destroy (query);
-    }
-
-    if (version < 1) {
-	Xapian::TermIterator t, t_end;
-
-	t_end = notmuch->xapian_db->allterms_end ("XTIMESTAMP");
-
-	for (t = notmuch->xapian_db->allterms_begin ("XTIMESTAMP");
-	     t != t_end;
-	     t++)
-	{
-	    Xapian::PostingIterator p, p_end;
-	    std::string term = *t;
-
-	    p_end = notmuch->xapian_db->postlist_end (term);
-
-	    for (p = notmuch->xapian_db->postlist_begin (term);
-		 p != p_end;
-		 p++)
-	    {
-		if (do_progress_notify) {
-		    progress_notify (closure, (double) count / total);
-		    do_progress_notify = 0;
-		}
-
-		db->delete_document (*p);
-	    }
-	}
-    }
+    db->commit_transaction ();
 
     if (timer_is_active) {
 	/* Now stop the timer. */
-- 
2.0.0

[PATCH v4 06/11] lib: Use database features to drive upgrade

[Austin Clements]

Previously, we had database version information hard-coded in the
upgrade code.  Slightly re-organize the upgrade process around the set
of new database features to be enabled by the upgrade.
---
 lib/database.cc | 24 +++++++++++++-----------
 1 file changed, 13 insertions(+), 11 deletions(-)

diff --git a/lib/database.cc b/lib/database.cc
index 1e46fc8..525c9c3 100644
--- a/lib/database.cc
+++ b/lib/database.cc
@@ -1202,11 +1202,12 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 			  void *closure)
 {
     void *local = talloc_new (NULL);
+    Xapian::TermIterator t, t_end;
     Xapian::WritableDatabase *db;
     struct sigaction action;
     struct itimerval timerval;
     notmuch_bool_t timer_is_active = FALSE;
-    unsigned int version;
+    enum _notmuch_features target_features, new_features;
     notmuch_status_t status;
     unsigned int count = 0, total = 0;
 
@@ -1216,9 +1217,10 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 
     db = static_cast <Xapian::WritableDatabase *> (notmuch->xapian_db);
 
-    version = notmuch_database_get_version (notmuch);
+    target_features = notmuch->features | NOTMUCH_FEATURES_CURRENT;
+    new_features = NOTMUCH_FEATURES_CURRENT & ~notmuch->features;
 
-    if (version >= NOTMUCH_DATABASE_VERSION)
+    if (! new_features)
 	return NOTMUCH_STATUS_SUCCESS;
 
     if (progress_notify) {
@@ -1244,18 +1246,17 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 
     /* Set the target features so we write out changes in the desired
      * format. */
-    notmuch->features |= NOTMUCH_FEATURES_CURRENT;
+    notmuch->features = target_features;
 
     /* Before version 1, each message document had its filename in the
      * data field. Copy that into the new format by calling
      * notmuch_message_add_filename.
      */
-    if (version < 1) {
+    if (new_features & NOTMUCH_FEATURE_FILE_TERMS) {
 	notmuch_query_t *query = notmuch_query_create (notmuch, "");
 	notmuch_messages_t *messages;
 	notmuch_message_t *message;
 	char *filename;
-	Xapian::TermIterator t, t_end;
 
 	total = notmuch_query_count_messages (query);
 
@@ -1284,11 +1285,12 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 	}
 
 	notmuch_query_destroy (query);
+    }
 
-	/* Also, before version 1 we stored directory timestamps in
-	 * XTIMESTAMP documents instead of the current XDIRECTORY
-	 * documents. So copy those as well. */
-
+    /* Also, before version 1 we stored directory timestamps in
+     * XTIMESTAMP documents instead of the current XDIRECTORY
+     * documents. So copy those as well. */
+    if (new_features & NOTMUCH_FEATURE_DIRECTORY_DOCS) {
 	t_end = notmuch->xapian_db->allterms_end ("XTIMESTAMP");
 
 	for (t = notmuch->xapian_db->allterms_begin ("XTIMESTAMP");
@@ -1332,7 +1334,7 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
      * stemmed. Change it to the current boolean prefix. Add "path:"
      * prefixes while at it.
      */
-    if (version < 2) {
+    if (new_features & NOTMUCH_FEATURE_BOOL_FOLDER) {
 	notmuch_query_t *query = notmuch_query_create (notmuch, "");
 	notmuch_messages_t *messages;
 	notmuch_message_t *message;
-- 
2.0.0

[PATCH v4 07/11] lib: Reorganize upgrade around document types

[Austin Clements]

Rather than potentially making multiple passes over the same type of
data in the database, reorganize upgrade around each type of data that
may be upgraded.  This eliminates code duplication, will make
multi-version upgrades faster, and will let us improve progress
reporting.
---
 lib/database.cc | 72 +++++++++++++++++++++------------------------------------
 1 file changed, 26 insertions(+), 46 deletions(-)

diff --git a/lib/database.cc b/lib/database.cc
index 525c9c3..6a7ce29 100644
--- a/lib/database.cc
+++ b/lib/database.cc
@@ -1248,11 +1248,9 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
      * format. */
     notmuch->features = target_features;
 
-    /* Before version 1, each message document had its filename in the
-     * data field. Copy that into the new format by calling
-     * notmuch_message_add_filename.
-     */
-    if (new_features & NOTMUCH_FEATURE_FILE_TERMS) {
+    /* Perform per-message upgrades. */
+    if (new_features &
+	(NOTMUCH_FEATURE_FILE_TERMS | NOTMUCH_FEATURE_BOOL_FOLDER)) {
 	notmuch_query_t *query = notmuch_query_create (notmuch, "");
 	notmuch_messages_t *messages;
 	notmuch_message_t *message;
@@ -1271,13 +1269,27 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 
 	    message = notmuch_messages_get (messages);
 
-	    filename = _notmuch_message_talloc_copy_data (message);
-	    if (filename && *filename != '\0') {
-		_notmuch_message_add_filename (message, filename);
-		_notmuch_message_clear_data (message);
-		_notmuch_message_sync (message);
+	    /* Before version 1, each message document had its
+	     * filename in the data field. Copy that into the new
+	     * format by calling notmuch_message_add_filename.
+	     */
+	    if (new_features & NOTMUCH_FEATURE_FILE_TERMS) {
+		filename = _notmuch_message_talloc_copy_data (message);
+		if (filename && *filename != '\0') {
+		    _notmuch_message_add_filename (message, filename);
+		    _notmuch_message_clear_data (message);
+		}
+		talloc_free (filename);
 	    }
-	    talloc_free (filename);
+
+	    /* Prior to version 2, the "folder:" prefix was
+	     * probabilistic and stemmed. Change it to the current
+	     * boolean prefix. Add "path:" prefixes while at it.
+	     */
+	    if (new_features & NOTMUCH_FEATURE_BOOL_FOLDER)
+		_notmuch_message_upgrade_folder (message);
+
+	    _notmuch_message_sync (message);
 
 	    notmuch_message_destroy (message);
 
@@ -1287,7 +1299,9 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 	notmuch_query_destroy (query);
     }
 
-    /* Also, before version 1 we stored directory timestamps in
+    /* Perform per-directory upgrades. */
+
+    /* Before version 1 we stored directory timestamps in
      * XTIMESTAMP documents instead of the current XDIRECTORY
      * documents. So copy those as well. */
     if (new_features & NOTMUCH_FEATURE_DIRECTORY_DOCS) {
@@ -1329,40 +1343,6 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 	}
     }
 
-    /*
-     * Prior to version 2, the "folder:" prefix was probabilistic and
-     * stemmed. Change it to the current boolean prefix. Add "path:"
-     * prefixes while at it.
-     */
-    if (new_features & NOTMUCH_FEATURE_BOOL_FOLDER) {
-	notmuch_query_t *query = notmuch_query_create (notmuch, "");
-	notmuch_messages_t *messages;
-	notmuch_message_t *message;
-
-	count = 0;
-	total = notmuch_query_count_messages (query);
-
-	for (messages = notmuch_query_search_messages (query);
-	     notmuch_messages_valid (messages);
-	     notmuch_messages_move_to_next (messages)) {
-	    if (do_progress_notify) {
-		progress_notify (closure, (double) count / total);
-		do_progress_notify = 0;
-	    }
-
-	    message = notmuch_messages_get (messages);
-
-	    _notmuch_message_upgrade_folder (message);
-	    _notmuch_message_sync (message);
-
-	    notmuch_message_destroy (message);
-
-	    count++;
-	}
-
-	notmuch_query_destroy (query);
-    }
-
     db->set_metadata ("features", _print_features (local, notmuch->features));
     db->set_metadata ("version", STRINGIFY (NOTMUCH_DATABASE_VERSION));
 
-- 
2.0.0

[PATCH v4 08/11] lib: Report progress for combined upgrade operation

[Austin Clements]

Previously, some parts of upgrade didn't report progress and for
others it was possible for the progress meter to restart at 0 part way
through the upgrade because each stage was reported separately.

Fix this by computing the total amount of work that needs to be done
up-front and updating completed work monotonically.
---
 lib/database.cc | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/lib/database.cc b/lib/database.cc
index 6a7ce29..53397bb 100644
--- a/lib/database.cc
+++ b/lib/database.cc
@@ -1241,6 +1241,19 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 	timer_is_active = TRUE;
     }
 
+    /* Figure out how much total work we need to do. */
+    if (new_features &
+	(NOTMUCH_FEATURE_FILE_TERMS | NOTMUCH_FEATURE_BOOL_FOLDER)) {
+	notmuch_query_t *query = notmuch_query_create (notmuch, "");
+	total += notmuch_query_count_messages (query);
+	notmuch_query_destroy (query);
+    }
+    if (new_features & NOTMUCH_FEATURE_DIRECTORY_DOCS) {
+	t_end = db->allterms_end ("XTIMESTAMP");
+	for (t = db->allterms_begin ("XTIMESTAMP"); t != t_end; t++)
+	    ++total;
+    }
+
     /* Perform the upgrade in a transaction. */
     db->begin_transaction (true);
 
@@ -1256,8 +1269,6 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 	notmuch_message_t *message;
 	char *filename;
 
-	total = notmuch_query_count_messages (query);
-
 	for (messages = notmuch_query_search_messages (query);
 	     notmuch_messages_valid (messages);
 	     notmuch_messages_move_to_next (messages))
@@ -1340,6 +1351,8 @@ notmuch_database_upgrade (notmuch_database_t *notmuch,
 
 		db->delete_document (*p);
 	    }
+
+	    ++count;
 	}
     }
 
-- 
2.0.0

[PATCH v4 09/11] lib: Support empty header values in database

[Austin Clements]

Commit 567bcbc2 introduced support for storing various headers in
document values.  However, doing so in a backwards-compatible way
meant that genuinely empty header values could not be distinguished
from the old behavior of not storing the headers at all, so these
required parsing the original message.

Now that we have database features, new databases can declare that all
messages have header values, so if we have this feature flag, we can
use the stored header value even if it's the empty string.

This requires slight cleanup to notmuch_message_get_header, since the
code previously couldn't distinguish between empty headers and headers
that are never stored in the database (previously this distinction
didn't matter).
---
 lib/message.cc | 45 +++++++++++++++++++++++++++------------------
 1 file changed, 27 insertions(+), 18 deletions(-)

diff --git a/lib/message.cc b/lib/message.cc
index 3f93426..ed8c59e 100644
--- a/lib/message.cc
+++ b/lib/message.cc
@@ -414,26 +414,35 @@ _notmuch_message_ensure_message_file (notmuch_message_t *message)
 const char *
 notmuch_message_get_header (notmuch_message_t *message, const char *header)
 {
-    try {
-	    std::string value;
-
-	    /* Fetch header from the appropriate xapian value field if
-	     * available */
-	    if (strcasecmp (header, "from") == 0)
-		value = message->doc.get_value (NOTMUCH_VALUE_FROM);
-	    else if (strcasecmp (header, "subject") == 0)
-		value = message->doc.get_value (NOTMUCH_VALUE_SUBJECT);
-	    else if (strcasecmp (header, "message-id") == 0)
-		value = message->doc.get_value (NOTMUCH_VALUE_MESSAGE_ID);
-
-	    if (!value.empty())
+    Xapian::valueno slot = Xapian::BAD_VALUENO;
+
+    /* Fetch header from the appropriate xapian value field if
+     * available */
+    if (strcasecmp (header, "from") == 0)
+	slot = NOTMUCH_VALUE_FROM;
+    else if (strcasecmp (header, "subject") == 0)
+	slot = NOTMUCH_VALUE_SUBJECT;
+    else if (strcasecmp (header, "message-id") == 0)
+	slot = NOTMUCH_VALUE_MESSAGE_ID;
+
+    if (slot != Xapian::BAD_VALUENO) {
+	try {
+	    std::string value = message->doc.get_value (slot);
+
+	    /* If we have NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES, then
+	     * empty values indicate empty headers.  If we don't, then
+	     * it could just mean we didn't record the header. */
+	    if ((message->notmuch->features &
+		 NOTMUCH_FEATURE_FROM_SUBJECT_ID_VALUES) ||
+		! value.empty())
 		return talloc_strdup (message, value.c_str ());
 
-    } catch (Xapian::Error &error) {
-	fprintf (stderr, "A Xapian exception occurred when reading header: %s\n",
-		 error.get_msg().c_str());
-	message->notmuch->exception_reported = TRUE;
-	return NULL;
+	} catch (Xapian::Error &error) {
+	    fprintf (stderr, "A Xapian exception occurred when reading header: %s\n",
+		     error.get_msg().c_str());
+	    message->notmuch->exception_reported = TRUE;
+	    return NULL;
+	}
     }
 
     /* Otherwise fall back to parsing the file */
-- 
2.0.0

[PATCH v4 10/11] lib: Return an error from operations that require an upgrade

[Austin Clements]

Previously, there was no protection against a caller invoking an
operation on an old database version that would effectively corrupt
the database by treating it like a newer version.

According to notmuch.h, any caller that opens the database in
read/write mode is supposed to check if the database needs upgrading
and perform an upgrade if it does.  This would protect against this,
but nobody (even the CLI) actually does this.

However, with features, it's easy to protect against incompatible
operations on a fine-grained basis.  This lightweight change allows
callers to safely operate on old database versions, while preventing
specific operations that would corrupt the database with an
informative error message.
---
 lib/database.cc  |  5 +++++
 lib/directory.cc |  5 +++++
 lib/message.cc   |  8 ++++++++
 lib/notmuch.h    | 16 ++++++++++++++++
 4 files changed, 34 insertions(+)

diff --git a/lib/database.cc b/lib/database.cc
index 53397bb..5116188 100644
--- a/lib/database.cc
+++ b/lib/database.cc
@@ -316,6 +316,8 @@ notmuch_status_to_string (notmuch_status_t status)
 	return "Unbalanced number of calls to notmuch_database_begin_atomic/end_atomic";
     case NOTMUCH_STATUS_UNSUPPORTED_OPERATION:
 	return "Unsupported operation";
+    case NOTMUCH_STATUS_UPGRADE_REQUIRED:
+	return "Operation requires a database upgrade";
     default:
     case NOTMUCH_STATUS_LAST_STATUS:
 	return "Unknown error status value";
@@ -2226,6 +2228,9 @@ notmuch_database_find_message_by_filename (notmuch_database_t *notmuch,
     if (message_ret == NULL)
 	return NOTMUCH_STATUS_NULL_POINTER;
 
+    if (! (notmuch->features & NOTMUCH_FEATURE_FILE_TERMS))
+	return NOTMUCH_STATUS_UPGRADE_REQUIRED;
+
     /* return NULL on any failure */
     *message_ret = NULL;
 
diff --git a/lib/directory.cc b/lib/directory.cc
index 6a3ffed..8daaec8 100644
--- a/lib/directory.cc
+++ b/lib/directory.cc
@@ -105,6 +105,11 @@ _notmuch_directory_create (notmuch_database_t *notmuch,
     const char *db_path;
     notmuch_bool_t create = (flags & NOTMUCH_FIND_CREATE);
 
+    if (! (notmuch->features & NOTMUCH_FEATURE_DIRECTORY_DOCS)) {
+	*status_ret = NOTMUCH_STATUS_UPGRADE_REQUIRED;
+	return NULL;
+    }
+
     *status_ret = NOTMUCH_STATUS_SUCCESS;
 
     path = _notmuch_database_relative_path (notmuch, path);
diff --git a/lib/message.cc b/lib/message.cc
index ed8c59e..68f7e68 100644
--- a/lib/message.cc
+++ b/lib/message.cc
@@ -655,6 +655,10 @@ _notmuch_message_add_filename (notmuch_message_t *message,
     if (filename == NULL)
 	INTERNAL_ERROR ("Message filename cannot be NULL.");
 
+    if (! (message->notmuch->features & NOTMUCH_FEATURE_FILE_TERMS) ||
+	! (message->notmuch->features & NOTMUCH_FEATURE_BOOL_FOLDER))
+	return NOTMUCH_STATUS_UPGRADE_REQUIRED;
+
     relative = _notmuch_database_relative_path (message->notmuch, filename);
 
     status = _notmuch_database_split_path (local, relative, &directory, NULL);
@@ -699,6 +703,10 @@ _notmuch_message_remove_filename (notmuch_message_t *message,
     notmuch_private_status_t private_status;
     notmuch_status_t status;
 
+    if (! (message->notmuch->features & NOTMUCH_FEATURE_FILE_TERMS) ||
+	! (message->notmuch->features & NOTMUCH_FEATURE_BOOL_FOLDER))
+	return NOTMUCH_STATUS_UPGRADE_REQUIRED;
+
     status = _notmuch_database_filename_to_direntry (
 	local, message->notmuch, filename, NOTMUCH_FIND_LOOKUP, &direntry);
     if (status || !direntry)
diff --git a/lib/notmuch.h b/lib/notmuch.h
index 3c5ec98..cbf2ba5 100644
--- a/lib/notmuch.h
+++ b/lib/notmuch.h
@@ -160,6 +160,10 @@ typedef enum _notmuch_status {
      */
     NOTMUCH_STATUS_UNSUPPORTED_OPERATION,
     /**
+     * The operation requires a database upgrade.
+     */
+    NOTMUCH_STATUS_UPGRADE_REQUIRED,
+    /**
      * Not an actual status value. Just a way to find out how many
      * valid status values there are.
      */
@@ -438,6 +442,9 @@ notmuch_database_end_atomic (notmuch_database_t *notmuch);
  *
  * NOTMUCH_STATUS_XAPIAN_EXCEPTION: A Xapian exception occurred;
  *	directory not retrieved.
+ *
+ * NOTMUCH_STATUS_UPGRADE_REQUIRED: The caller must upgrade the
+ * 	database to use this function.
  */
 notmuch_status_t
 notmuch_database_get_directory (notmuch_database_t *database,
@@ -490,6 +497,9 @@ notmuch_database_get_directory (notmuch_database_t *database,
  *
  * NOTMUCH_STATUS_READ_ONLY_DATABASE: Database was opened in read-only
  *	mode so no message can be added.
+ *
+ * NOTMUCH_STATUS_UPGRADE_REQUIRED: The caller must upgrade the
+ * 	database to use this function.
  */
 notmuch_status_t
 notmuch_database_add_message (notmuch_database_t *database,
@@ -520,6 +530,9 @@ notmuch_database_add_message (notmuch_database_t *database,
  *
  * NOTMUCH_STATUS_READ_ONLY_DATABASE: Database was opened in read-only
  *	mode so no message can be removed.
+ *
+ * NOTMUCH_STATUS_UPGRADE_REQUIRED: The caller must upgrade the
+ * 	database to use this function.
  */
 notmuch_status_t
 notmuch_database_remove_message (notmuch_database_t *database,
@@ -575,6 +588,9 @@ notmuch_database_find_message (notmuch_database_t *database,
  * NOTMUCH_STATUS_OUT_OF_MEMORY: Out of memory, creating the message object
  *
  * NOTMUCH_STATUS_XAPIAN_EXCEPTION: A Xapian exception occurred
+ *
+ * NOTMUCH_STATUS_UPGRADE_REQUIRED: The caller must upgrade the
+ * 	database to use this function.
  */
 notmuch_status_t
 notmuch_database_find_message_by_filename (notmuch_database_t *notmuch,
-- 
2.0.0

[PATCH v4 11/11] lib: Update doc of notmuch_database_{needs_upgrade, upgrade}

[Austin Clements]

Clients are no longer required to call these functions after opening a
database in read/write mode (which is good, because almost none of
them do!).
---
 lib/notmuch.h | 23 ++++++++++++++---------
 1 file changed, 14 insertions(+), 9 deletions(-)

diff --git a/lib/notmuch.h b/lib/notmuch.h
index cbf2ba5..21a5225 100644
--- a/lib/notmuch.h
+++ b/lib/notmuch.h
@@ -352,22 +352,27 @@ unsigned int
 notmuch_database_get_version (notmuch_database_t *database);
 
 /**
- * Does this database need to be upgraded before writing to it?
+ * Can the database be upgraded to a newer database version?
  *
- * If this function returns TRUE then no functions that modify the
- * database (notmuch_database_add_message, notmuch_message_add_tag,
- * notmuch_directory_set_mtime, etc.) will work unless the function
- * notmuch_database_upgrade is called successfully first.
+ * If this function returns TRUE, then the caller may call
+ * notmuch_database_upgrade to upgrade the database.  If the caller
+ * does not upgrade an out-of-date database, then some functions may
+ * fail with NOTMUCH_STATUS_UPGRADE_REQUIRED.  This always returns
+ * FALSE for a read-only database because there's no way to upgrade a
+ * read-only database.
  */
 notmuch_bool_t
 notmuch_database_needs_upgrade (notmuch_database_t *database);
 
 /**
- * Upgrade the current database.
+ * Upgrade the current database to the latest supported version.
  *
- * After opening a database in read-write mode, the client should
- * check if an upgrade is needed (notmuch_database_needs_upgrade) and
- * if so, upgrade with this function before making any modifications.
+ * This ensures that all current notmuch functionality will be
+ * available on the database.  After opening a database in read-write
+ * mode, it is recommended that clients check if an upgrade is needed
+ * (notmuch_database_needs_upgrade) and if so, upgrade with this
+ * function before making any modifications.  If
+ * notmuch_database_needs_upgrade returns FALSE, this will be a no-op.
  *
  * The optional progress_notify callback can be used by the caller to
  * provide progress indication to the user. If non-NULL it will be
-- 
2.0.0

Re: [PATCH v4 00/11] Implement and use database "features"

[David Bremner]

[-- Attachment #1: Type: text/plain, Size: 546 bytes --]

Austin Clements <amdragon@mit.edu> writes:

> This is v4 of id:1406859003-11561-1-git-send-email-amdragon@mit.edu.
> In addition to rebasing to current master, this makes several tidying
> changes: it gives the features enum a name for better
> self-documentation and type-safety; improves the robustness of
> _parse_features to NULL pointers; requests upgrades if the database
> version is old, even if it supports all current features; improves
> various comments; and fixes some style errors.
>

pushed this version of the series to master

d

[-- Attachment #2: Type: application/pgp-signature, Size: 647 bytes --]