Patchwork [V3] obsolete: use C code for headrevs calculation

login
register
mail settings
Submitter Durham Goode
Date Sept. 18, 2014, 2:01 a.m.
Message ID <4c581ed8d04c771305f3.1411005661@dev2000.prn2.facebook.com>
Download mbox | patch
Permalink /patch/5863/
State Superseded
Headers show

Comments

Durham Goode - Sept. 18, 2014, 2:01 a.m.
# HG changeset patch
# User Durham Goode <durham@fb.com>
# Date 1410908601 25200
#      Tue Sep 16 16:03:21 2014 -0700
# Node ID 4c581ed8d04c771305f388385a33d993d4c696ae
# Parent  48791c2bea1ceda4e4f28bc11651e281d636ce1a
obsolete: use C code for headrevs calculation

Previously, if there were filtered revs the repository could not use the C fast
path for computing the head revs in the changelog. This slowed down many
operations in large repositories.

This adds the ability to filter revs to the C fast path. This speeds up histedit
on repositories with filtered revs by 30% (13s to 9s). This could be improved
further by sorting the filtered revs and walking the sorted list while we walk
the changelog, but even this initial version that just calls __contains__ is
still massively faster.

The new C api is compatible for both new and old python clients, and the new
python client can call both new and old C apis.
Antoine Pitrou - Sept. 18, 2014, 11:24 a.m.
On Wed, 17 Sep 2014 19:01:01 -0700
Durham Goode <durham@fb.com> wrote:
>  
> -static PyObject *index_headrevs(indexObject *self)
> +static int check_filter(PyObject *filter, Py_ssize_t arg, int *isfiltered) {

That's complicated. Why doesn't your function return either 1, 0 or -1?
By the way, PyObject_IsTrue() does exactly that :-)
(meaning you should check its error return as well, or return it
directly)
Durham Goode - Sept. 18, 2014, 6:13 p.m.
On 9/18/14, 4:24 AM, Antoine Pitrou wrote:
> On Wed, 17 Sep 2014 19:01:01 -0700
> Durham Goode <durham@fb.com> wrote:
>>   
>> -static PyObject *index_headrevs(indexObject *self)
>> +static int check_filter(PyObject *filter, Py_ssize_t arg, int *isfiltered) {
> That's complicated. Why doesn't your function return either 1, 0 or -1?
> By the way, PyObject_IsTrue() does exactly that :-)
> (meaning you should check its error return as well, or return it
> directly)
Overloading the return value to mean multiple things seemed odd. But if 
that's a standard pattern, I'll do that.

Patch

diff --git a/mercurial/changelog.py b/mercurial/changelog.py
--- a/mercurial/changelog.py
+++ b/mercurial/changelog.py
@@ -171,8 +171,13 @@ 
 
     def headrevs(self):
         if self.filteredrevs:
-            # XXX we should fix and use the C version
-            return self._headrevs()
+            try:
+                return self.index.headrevs(self.filteredrevs)
+            # AttributeError covers non-c-extension environments.
+            # TypeError allows us work with old c extensions.
+            except (AttributeError, TypeError):
+                return self._headrevs()
+
         return super(changelog, self).headrevs()
 
     def strip(self, *args, **kwargs):
diff --git a/mercurial/parsers.c b/mercurial/parsers.c
--- a/mercurial/parsers.c
+++ b/mercurial/parsers.c
@@ -508,6 +508,7 @@ 
 	Py_ssize_t length;     /* current number of elements */
 	PyObject *added;       /* populated on demand */
 	PyObject *headrevs;    /* cache, invalidated on changes */
+	PyObject *filteredrevs;/* filtered revs set */
 	nodetree *nt;          /* base-16 trie */
 	int ntlength;          /* # nodes in use */
 	int ntcapacity;        /* # nodes allocated */
@@ -823,15 +824,59 @@ 
 	return newlist;
 }
 
-static PyObject *index_headrevs(indexObject *self)
+static int check_filter(PyObject *filter, Py_ssize_t arg, int *isfiltered) {
+	if (filter) {
+		PyObject *arglist = Py_BuildValue("(n)", arg);
+		if (!arglist) {
+			return -1;
+		}
+
+		PyObject *result = PyEval_CallObject(filter, arglist);
+		Py_DECREF(arglist);
+		if (!result) {
+			return -1;
+		}
+
+		if (PyObject_IsTrue(result)) {
+			*isfiltered = 1;
+		} else {
+			*isfiltered = 0;
+		}
+		Py_DECREF(result);
+	} else {
+		*isfiltered = 0;
+	}
+	return 0;
+}
+
+static PyObject *index_headrevs(indexObject *self, PyObject *args)
 {
 	Py_ssize_t i, len, addlen;
 	char *nothead = NULL;
 	PyObject *heads;
+	PyObject *filter = NULL;
+	PyObject *filteredrevs = Py_None;
 
-	if (self->headrevs)
+	if (!PyArg_ParseTuple(args, "|O", &filteredrevs)) {
+		return NULL;
+	}
+
+	if (self->headrevs && filteredrevs == self->filteredrevs)
 		return list_copy(self->headrevs);
 
+	Py_DECREF(self->filteredrevs);
+	self->filteredrevs = filteredrevs;
+	Py_INCREF(filteredrevs);
+
+	if (filteredrevs != Py_None) {
+		filter = PyObject_GetAttrString(filteredrevs, "__contains__");
+		if (!filter) {
+			PyErr_SetString(PyExc_TypeError,
+				"filteredrevs has no attribute __contains__");
+			goto bail;
+		}
+	}
+
 	len = index_length(self) - 1;
 	heads = PyList_New(0);
 	if (heads == NULL)
@@ -850,9 +895,24 @@ 
 		goto bail;
 
 	for (i = 0; i < self->raw_length; i++) {
-		const char *data = index_deref(self, i);
-		int parent_1 = getbe32(data + 24);
-		int parent_2 = getbe32(data + 28);
+		const char *data;
+		int parent_1, parent_2, isfiltered;
+
+		if (check_filter(filter, i, &isfiltered)) {
+			PyErr_SetString(PyExc_TypeError,
+				"unable to check filter");
+			goto bail;
+		}
+
+		if (isfiltered) {
+			nothead[i] = 1;
+			continue;
+		}
+
+		data = index_deref(self, i);
+		parent_1 = getbe32(data + 24);
+		parent_2 = getbe32(data + 28);
+
 		if (parent_1 >= 0)
 			nothead[parent_1] = 1;
 		if (parent_2 >= 0)
@@ -866,12 +926,25 @@ 
 		PyObject *p1 = PyTuple_GET_ITEM(rev, 5);
 		PyObject *p2 = PyTuple_GET_ITEM(rev, 6);
 		long parent_1, parent_2;
+		int isfiltered;
 
 		if (!PyInt_Check(p1) || !PyInt_Check(p2)) {
 			PyErr_SetString(PyExc_TypeError,
 					"revlog parents are invalid");
 			goto bail;
 		}
+
+		if (check_filter(filter, i, &isfiltered)) {
+			PyErr_SetString(PyExc_TypeError,
+				"unable to check filter");
+			goto bail;
+		}
+
+		if (isfiltered) {
+			nothead[i] = 1;
+			continue;
+		}
+
 		parent_1 = PyInt_AS_LONG(p1);
 		parent_2 = PyInt_AS_LONG(p2);
 		if (parent_1 >= 0)
@@ -894,9 +967,11 @@ 
 
 done:
 	self->headrevs = heads;
+	Py_XDECREF(filter);
 	free(nothead);
 	return list_copy(self->headrevs);
 bail:
+	Py_XDECREF(filter);
 	Py_XDECREF(heads);
 	free(nothead);
 	return NULL;
@@ -1896,6 +1971,8 @@ 
 	self->cache = NULL;
 	self->data = NULL;
 	self->headrevs = NULL;
+	self->filteredrevs = Py_None;
+	Py_INCREF(Py_None);
 	self->nt = NULL;
 	self->offsets = NULL;
 
@@ -1945,6 +2022,7 @@ 
 static void index_dealloc(indexObject *self)
 {
 	_index_clearcaches(self);
+	Py_XDECREF(self->filteredrevs);
 	Py_XDECREF(self->data);
 	Py_XDECREF(self->added);
 	PyObject_Del(self);
@@ -1977,7 +2055,7 @@ 
 	 "clear the index caches"},
 	{"get", (PyCFunction)index_m_get, METH_VARARGS,
 	 "get an index entry"},
-	{"headrevs", (PyCFunction)index_headrevs, METH_NOARGS,
+	{"headrevs", (PyCFunction)index_headrevs, METH_VARARGS,
 	 "get head revisions"},
 	{"insert", (PyCFunction)index_insert, METH_VARARGS,
 	 "insert an index entry"},