Skip to content

Field aliases now use hashes, not linear scan #141

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 4 commits into
base: master
Choose a base branch
from
Open

Field aliases now use hashes, not linear scan #141

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
de0ced9
Update PHPUnit to version 7 (old version 4.9 wasnt supported by PHP 7…
bigbes Jul 1, 2018
d90237b
Use packpack instead of obsolete tarantool/build in .travis.yml
bigbes Aug 14, 2018
d62eb1a
Update travis.yml
bigbes Nov 4, 2018
00c81ba
Replace linear space scan with hash for update fields
bigbes Jul 1, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
138 changes: 74 additions & 64 deletions src/tarantool_schema.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,33 +14,53 @@
#define MUR_SEED 13
#include "third_party/msgpuck.h"

int mh_indexcmp_eq(
const struct schema_index_value **lval,
const struct schema_index_value **rval,
void *arg
) {
(void )arg;
if ((*lval)->key.id_len != (*rval)->key.id_len)
return 0;
return !memcmp((*lval)->key.id, (*rval)->key.id, (*rval)->key.id_len);
}
#define MH_DEFINE_CMPFUNC(NAME, TYPE) \
int mh_##NAME##cmp_eq(const TYPE **lval, const TYPE **rval, \
void *arg) { \
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Double whitespace.

(void *) arg; \
if ((*lval)->key.id_len != (*rval)->key.id_len) \
return 0; \
return !memcmp((*lval)->key.id, (*rval)->key.id, \
(*rval)->key.id_len); \
} \
\
int mh_##NAME##cmp_key_eq(const struct schema_key *key, \
const TYPE **val, void *arg) { \
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Two whitespaces before tabs.

(void *) arg; \
if (key->id_len != (*val)->key.id_len) \
return 0; \
return !memcmp(key->id, (*val)->key.id, key->id_len); \
}

int mh_indexcmp_key_eq(
const struct schema_key *key,
const struct schema_index_value **val,
void *arg
) {
(void )arg;
if (key->id_len != (*val)->key.id_len)
return 0;
return !memcmp(key->id, (*val)->key.id, key->id_len);
}
MH_DEFINE_CMPFUNC(field, struct schema_field_value);
MH_DEFINE_CMPFUNC(index, struct schema_index_value);
MH_DEFINE_CMPFUNC(space, struct schema_space_value);
#undef MH_DEFINE_CMPFUNC

#define mh_arg_t void *

#define mh_eq(a, b, arg) mh_fieldcmp_eq(a, b, arg)
#define mh_eq_key(a, b, arg) mh_fieldcmp_key_eq(a, b, arg)
#define mh_hash(x, arg) PMurHash32(MUR_SEED, (*x)->key.id, \
(*x)->key.id_len)
#define mh_hash_key(x, arg) PMurHash32(MUR_SEED, (x)->id, (x)->id_len);

#define mh_node_t struct schema_field_value *
#define mh_key_t struct schema_key *

#define MH_CALLOC(x, y) pecalloc((x), (y), 1)
#define MH_FREE(x) pefree((x), 1)

#define mh_name _schema_field
#define MH_SOURCE 1
#include "third_party/mhash.h"

#define mh_arg_t void *

#define mh_eq(a, b, arg) mh_indexcmp_eq(a, b, arg)
#define mh_eq_key(a, b, arg) mh_indexcmp_key_eq(a, b, arg)
#define mh_hash(x, arg) PMurHash32(MUR_SEED, (*x)->key.id, (*x)->key.id_len)
#define mh_hash(x, arg) PMurHash32(MUR_SEED, (*x)->key.id, \
(*x)->key.id_len)
#define mh_hash_key(x, arg) PMurHash32(MUR_SEED, (x)->id, (x)->id_len);

#define mh_node_t struct schema_index_value *
Expand Down Expand Up @@ -79,7 +99,7 @@ schema_index_free(struct mh_schema_index_t *schema) {
do {
struct schema_key key_number = {
(void *)&(ivalue->index_number),
sizeof(uint32_t), 0
sizeof(ivalue->index_number), 0
};
index_slot = mh_schema_index_find(schema, &key_number,
NULL);
Expand All @@ -106,33 +126,12 @@ schema_index_free(struct mh_schema_index_t *schema) {
}
}

int
mh_spacecmp_eq(
const struct schema_space_value **lval,
const struct schema_space_value **rval,
void *arg) {
(void )arg;
if ((*lval)->key.id_len != (*rval)->key.id_len)
return 0;
return !memcmp((*lval)->key.id, (*rval)->key.id, (*rval)->key.id_len);
}

int
mh_spacecmp_key_eq(
const struct schema_key *key,
const struct schema_space_value **val,
void *arg) {
(void )arg;
if (key->id_len != (*val)->key.id_len)
return 0;
return !memcmp(key->id, (*val)->key.id, key->id_len);
}

#define mh_arg_t void *

#define mh_eq(a, b, arg) mh_spacecmp_eq(a, b, arg)
#define mh_eq_key(a, b, arg) mh_spacecmp_key_eq(a, b, arg)
#define mh_hash(x, arg) PMurHash32(MUR_SEED, (*x)->key.id, (*x)->key.id_len)
#define mh_hash(x, arg) PMurHash32(MUR_SEED, (*x)->key.id, \
(*x)->key.id_len)
#define mh_hash_key(x, arg) PMurHash32(MUR_SEED, x->id, x->id_len);

#define mh_node_t struct schema_space_value *
Expand All @@ -158,6 +157,9 @@ schema_space_value_free(const struct schema_space_value *val) {
schema_index_free(val->index_hash);
mh_schema_index_delete(val->index_hash);
}
if (val->schema_hash) {
mh_schema_field_delete(val->schema_hash);
}
}
}

Expand All @@ -166,7 +168,8 @@ schema_space_free(struct mh_schema_space_t *schema) {
int pos = 0;
mh_int_t space_slot = 0;
mh_foreach(schema, pos) {
const struct schema_space_value *svalue, *sv1 = NULL, *sv2 = NULL;
const struct schema_space_value *svalue;
const struct schema_space_value *sv1 = NULL, *sv2 = NULL;
svalue = *mh_schema_space_node(schema, pos);
do {
struct schema_key key_number = {
Expand Down Expand Up @@ -254,7 +257,8 @@ parse_schema_space_value(struct schema_space_value *space_string,
goto error;
int i = 0;
for (i = 0; i < fmt_len; ++i) {
struct schema_field_value *val = &(space_string->schema_list[i]);
struct schema_field_value *val =
&(space_string->schema_list[i]);
if (mp_typeof(**tuple) != MP_MAP)
goto error;
uint32_t arrsz = mp_decode_map(tuple);
Expand All @@ -263,6 +267,11 @@ parse_schema_space_value(struct schema_space_value *space_string,
goto error;
}
val->field_number = i;
val->key.id = val->field_name;
val->key.id_len = val->field_name_len;
mh_schema_field_put(space_string->schema_hash,
(const struct schema_field_value **)&val,
NULL, NULL);
}
return 0;
error:
Expand All @@ -289,7 +298,8 @@ parse_schema_index_value(struct schema_index_value *index_string,
goto error;
if (mp_decode_array(tuple) != 2)
goto error;
struct schema_field_value *val = &(index_string->index_parts[i]);
struct schema_field_value *val =
&(index_string->index_parts[i]);
if (mp_typeof(**tuple) != MP_UINT)
goto error;
val->field_number = mp_decode_uint(tuple);
Expand Down Expand Up @@ -367,6 +377,7 @@ schema_add_space(
* }
*/
case 6:
space_string->schema_hash = mh_schema_field_new();
if (parse_schema_space_value(space_string, &tuple) < 0)
goto error;
break;
Expand Down Expand Up @@ -590,27 +601,26 @@ tarantool_schema_get_fid_by_string(
struct tarantool_schema *schema_obj, uint32_t sid,
const char *field_name, uint32_t field_name_len
) {
struct mh_schema_space_t *schema = schema_obj->space_hash;
struct schema_key space_key = {
(void *)&sid,
sizeof(uint32_t), 0
};
mh_int_t space_slot = mh_schema_space_find(schema, &space_key, NULL);
if (space_slot == mh_end(schema))
struct mh_schema_space_t *sschema = schema_obj->space_hash;
struct schema_key space_key = { (void *)&sid, sizeof(uint32_t), 0 };
mh_int_t space_slot = mh_schema_space_find(sschema, &space_key, NULL);
if (space_slot == mh_end(sschema))
return -1;
const struct schema_space_value *space = *mh_schema_space_node(schema,
const struct schema_space_value *space = *mh_schema_space_node(sschema,
space_slot);
int i = 0;
for (i = 0; i < space->schema_list_len; ++i) {
struct schema_field_value *val = &space->schema_list[i];
if (strncmp(val->field_name, field_name, field_name_len) == 0)
return val->field_number;
}
return -1;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It seems it was the singleton use of the schema_list/schema_list_len, so these fields can be removed now from struct schema_space_value?

struct schema_key field_key = { field_name, field_name_len, 0 };
struct mh_schema_field_t *fschema = space->schema_hash;
mh_int_t field_slot = mh_schema_field_find(fschema, &field_key, NULL);
if (field_slot == mh_end(fschema))
return -1;
const struct schema_field_value *field = *mh_schema_field_node(fschema,
field_slot);
return field->field_number;
}

struct tarantool_schema *tarantool_schema_new(int is_persistent) {
struct tarantool_schema *obj = pemalloc(sizeof(struct tarantool_schema *), 1);
struct tarantool_schema *obj = NULL;
obj = pemalloc(sizeof(struct tarantool_schema *), 1);
obj->space_hash = mh_schema_space_new();
return obj;
}
Expand Down
14 changes: 8 additions & 6 deletions src/tarantool_schema.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,17 +14,18 @@ enum field_type {
};

struct schema_field_value {
uint32_t field_number;
char *field_name;
uint32_t field_name_len;
enum field_type field_type;
struct schema_key key;
uint32_t field_number;
char *field_name;
uint32_t field_name_len;
enum field_type field_type;
};

struct schema_index_value {
struct schema_key key;
uint32_t index_number;
char *index_name;
uint32_t index_name_len;
uint32_t index_number;
struct schema_field_value *index_parts;
uint32_t index_parts_len;
};
Expand All @@ -33,10 +34,11 @@ struct mh_schema_index_t;

struct schema_space_value {
struct schema_key key;
uint32_t space_number;
char *space_name;
uint32_t space_name_len;
uint32_t space_number;
struct mh_schema_index_t *index_hash;
struct mh_schema_field_t *schema_hash;
struct schema_field_value *schema_list;
uint32_t schema_list_len;
};
Expand Down