[PATCH] tag: Automatically limit to messages whose tags will actually change.

[Austin Clements]

Quoth Jani Nikula on Nov 09 at  8:46 am:
> 
> FWIW, I reviewed this and didn't find any obvious problems. A few
> nitpicks below, though.
> 
> BR,
> Jani.
> 
> On Mon,  7 Nov 2011 22:55:23 -0500, Austin Clements <amdragon at MIT.EDU> wrote:
> > This optimizes the user's tagging query to exclude messages that won't
> > be affected by the tagging operation, saving computation and IO for
> > redundant tagging operations.
> > 
> > For example,
> >   notmuch tag +notmuch to:notmuch at notmuchmail.org
> > will now use the query
> >   ( to:notmuch at notmuchmail.org ) and (not tag:"notmuch")
> > 
> > In the past, we've often suggested that people do this exact
> > transformation by hand for slow tagging operations.  This makes that
> > unnecessary.
> > ---
> > I was about to implement this optimization in my initial tagging
> > script, but then I figured, why not just do it in notmuch so we can
> > stop telling people to do this by hand?
> > 
> >  NEWS          |    9 ++++++
> >  notmuch-tag.c |   76 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> >  2 files changed, 85 insertions(+), 0 deletions(-)
> > 
> > diff --git a/NEWS b/NEWS
> > index e00452a..9ca5e0c 100644
> > --- a/NEWS
> > +++ b/NEWS
> > @@ -16,6 +16,15 @@ Add search terms to  "notmuch dump"
> >    search/show/tag. The output file argument of dump is deprecated in
> >    favour of using stdout.
> >  
> > +Optimizations
> > +-------------
> > +
> > +Automatic tag query optimization
> > +
> > +  "notmuch tag" now automatically optimizes the user's query to
> > +  exclude messages whose tags won't change.  In the past, we've
> > +  suggested that people do this by hand; this is no longer necessary.
> > +
> >  Notmuch 0.9 (2011-10-01)
> >  ========================
> >  
> > diff --git a/notmuch-tag.c b/notmuch-tag.c
> > index dded39e..62c4bf1 100644
> > --- a/notmuch-tag.c
> > +++ b/notmuch-tag.c
> > @@ -30,6 +30,76 @@ handle_sigint (unused (int sig))
> >      interrupted = 1;
> >  }
> >  
> > +static char *
> > +_escape_tag (char *buf, const char *tag)
> > +{
> > +    const char *in = tag;
> > +    char *out = buf;
> > +    /* Boolean terms surrounded by double quotes can contain any
> > +     * character.  Double quotes are quoted by doubling them. */
> > +    *(out++) = '"';
> > +    while (*in) {
> > +	if (*in == '"')
> > +	    *(out++) = '"';
> > +	*(out++) = *(in++);
> > +    }
> > +    *(out++) = '"';
> 
> The parenthesis are unnecessary for *p++.

Removed.  I put these in out of paranoia, but I suppose it wouldn't be
an lvalue if it parsed differently.

> > +    *out = 0;
> > +    return buf;
> > +}
> > +
> > +static char *
> > +_optimize_tag_query (void *ctx, const char *orig_query_string, char *argv[],
> > +		     int *add_tags, int add_tags_count,
> > +		     int *remove_tags, int remove_tags_count)
> > +{
> > +    /* This is subtler than it looks.  Xapian ignores the '-' operator
> > +     * at the beginning both queries and parenthesized groups and,
> > +     * furthermore, the presence of a '-' operator at the beginning of
> > +     * a group can inhibit parsing of the previous operator.  Hence,
> > +     * the user-provided query MUST appear first, but it is safe to
> > +     * parenthesize and the exclusion part of the query must not use
> > +     * the '-' operator (though the NOT operator is fine). */
> > +
> > +    char *escaped, *query_string;
> > +    const char *join = "";
> > +    int i;
> > +    unsigned int max_tag_len = 0;
> > +
> > +    /* Allocate a buffer for escaping tags. */
> > +    for (i = 0; i < add_tags_count; i++)
> > +	if (strlen (argv[add_tags[i]] + 1) > max_tag_len)
> > +	    max_tag_len = strlen (argv[add_tags[i]] + 1);
> > +    for (i = 0; i < remove_tags_count; i++)
> > +	if (strlen (argv[remove_tags[i]] + 1) > max_tag_len)
> > +	    max_tag_len = strlen (argv[remove_tags[i]] + 1);
> > +    escaped = talloc_array(ctx, char, max_tag_len * 2 + 3);
> 
> Perhaps a comment here or above _escape_tag() explaining the worst case
> memory consumption of strlen(tag) * 2 + 3 for a tag of "s would be in
> order.

Definitely.  Done.

> It's unrelated, but looking at the above also made me check something
> I've suspected before: notmuch allows you to have empty or zero length
> tags "", which is probably not intentional.
> 
> There's no check for talloc failures here or below. But then there are
> few checks for that in the cli in general. *shrug*.

It's unfortunate that error handling obscures C code so much.  But
there's no sense in not handling errors, so I fixed this.

> > +
> > +    /* Build the new query string */
> > +    if (strcmp (orig_query_string, "*") == 0)
> > +	query_string = talloc_strdup (ctx, "(");
> > +    else
> > +	query_string = talloc_asprintf (ctx, "( %s ) and (", orig_query_string);
> > +
> > +    for (i = 0; i < add_tags_count; i++) {
> > +	query_string = talloc_asprintf_append_buffer (
> > +	    query_string, "%snot tag:%s", join,
> > +	    _escape_tag (escaped, argv[add_tags[i]] + 1));
> > +	join = " or ";
> > +    }
> > +    for (i = 0; i < remove_tags_count; i++) {
> > +	query_string = talloc_asprintf_append_buffer (
> > +	    query_string, "%stag:%s", join,
> > +	    _escape_tag (escaped, argv[remove_tags[i]] + 1));
> > +	join = " or ";
> > +    }
> > +
> > +    query_string = talloc_strdup_append_buffer (query_string, ")");
> > +
> > +    talloc_free (escaped);
> > +    return query_string;
> > +}
> > +
> >  int
> >  notmuch_tag_command (void *ctx, unused (int argc), unused (char *argv[]))
> >  {
> > @@ -93,6 +163,12 @@ notmuch_tag_command (void *ctx, unused (int argc), unused (char *argv[]))
> >  	return 1;
> >      }
> >  
> > +    /* Optimize the query so it excludes messages that already have
> > +     * the specified set of tags. */
> > +    query_string = _optimize_tag_query (ctx, query_string, argv,
> > +					add_tags, add_tags_count,
> > +					remove_tags, remove_tags_count);
> > +
> >      config = notmuch_config_open (ctx, NULL, NULL);
> >      if (config == NULL)
> >  	return 1;