greg/lmdb-erl
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Rename to lmdb #2

Merged
gburd merged 3 commits from gsb-rename-as-lmdb into master 2013-05-19 14:51:36 +00:00
Conversation 0 Commits 3 Files changed 16 +299 -1142
16 changed files with 299 additions and 1142 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

7
.gitignore vendored Normal file
View file

@ -0,0 +1,7 @@
*.beam
.eunit
ebin
c_src/*.o
deps/
priv/
*~

8
3rdLibs
View file

@ -1,8 +0,0 @@
o MDB: http://highlandsun.com/hyc/mdb/
source : from OpenLDAP git repository
commit : 7333b6bdc90c49c332d4728c62c7904e64c95077
license: OpenLDAP Public License
o UTHash: http://uthash.sourceforge.net/
version: 1.9.6
license: BSD revised

10
Makefile
View file

@ -1,6 +1,4 @@
# See LICENSE for licensing information.
MODULE = emdb
MODULE = lmdb
DIALYZER = dialyzer
REBAR = rebar
@ -56,10 +54,10 @@ compile-for-triq:
@$(REBAR) -D QC -D QC_TRIQ compile triq compile_only=true
plt: compile
@$(DIALYZER) --build_plt --output_plt .$(TARGET).plt -pa deps/lager/ebin --apps kernel stdlib
@$(DIALYZER) --build_plt --output_plt .$(TARGET).plt -pa --apps kernel stdlib
analyze: compile
@$(DIALYZER) --plt .$(TARGET).plt -pa deps/lager/ebin ebin
@$(DIALYZER) --plt .$(TARGET).plt
repl:
@$(ERL) -pa ebin -pz deps/lager/ebin
@$(ERL) exec erl -pa $PWD/ebin -pa +B

130
README.md
View file

@ -1,14 +1,28 @@
EMDB
====
An Erlang NIF for LMDB
======================
EMDB is a NIF library for the [Memory-Mapped Database](http://highlandsun.com/hyc/mdb/) database, aka. MDB.
This is an Erlang NIF for OpenLDAP's Lightning Memory-Mapped Database (LMDB) database library. LMDB is an ultra-fast, ultra-compact key-value data store developed by Symas for the OpenLDAP Project. It uses memory-mapped files, so it has the read performance of a pure in-memory database while still offering the persistence of standard disk-based databases, and is only limited to the size of the virtual address space, (it is not limited to the size of physical RAM). LMDB was originally called MDB, but was renamed to avoid confusion with other software associated with the name MDB
The main purpose of this package is to provide an Erlang API for this excellent BTREE implementation. Secondly to build an alternative Riak/KV [backend](http://wiki.basho.com/Storage-Backends.html) and Riak's AAE feature based on this. Finally it'd be nice to build an ETS-compatible API (ala "lets" for LevelDB) to ease adoption in other places where DETS is just not up to the task.
Quick Overview
--------------
LMDB is a tiny database with some excellent properties:
* Ordered-map interface (keys are always sorted, supports range lookups)
* Reader/writer transactions: readers don't block writers and writers don't block readers. Writers are fully serialized, so writes are always deadlock-free.
* Read transactions are extremely cheap, and can be performed using no mallocs or any other blocking calls.
* Supports multi-thread and multi-process concurrency, environments may be opened by multiple processes on the same host.
* Multiple sub-databases may be created with transactions covering all sub-databases.
* Memory-mapped, allowing for zero-copy lookup and iteration.
* Maintenance-free, no external process or background cleanup/compaction required.
* No application-level caching. LMDB fully exploits the operating system's buffer cache.
* 32KB of object code and 6KLOC of C.
The main purpose of this integration is to provide Erlang programmers access to this excellent, and open source friendly, BTREE implementation.
Requirements
------------
* Erlang R14B04+
* GCC 4.2+ or MS VisualStudio 2010+
* Clang, GCC 4.2+ or MS VisualStudio 2010+
Build
-----
@ -19,14 +33,12 @@ $ make
API
---
The following functions were implemented:
* `open/1`: equivalent to `emdb:open(DirName, 10485760)`.
* `open/2`: equivalent to `emdb:open(DirName, 10485760, 0)`.
* `open/1`: equivalent to `lmdb:open(DirName, 10485760)`.
* `open/2`: equivalent to `lmdb:open(DirName, 10485760, 0)`.
* `open/3`: creates a new MDB database. This call also re-open an already existing one. Arguments are:
* DirName: database directory name
* MapSize: database map size (see [map.hrl](http://gitorious.org/mdb/mdb/blobs/master/libraries/libmdb/mdb.h))
* EnvFlags: database environment flags (see [map.hrl](http://gitorious.org/mdb/mdb/blobs/master/libraries/libmdb/mdb.h)). The possible values are defined in **emdb.hrl**.
* EnvFlags: database environment flags (see [map.hrl](http://gitorious.org/mdb/mdb/blobs/master/libraries/libmdb/mdb.h)). The possible values are defined in **lmdb.hrl**.
* `close/2`: closes the database
* `put/2`: inserts Key with value Val into the database. Assumes that the key is not present, 'key_exit' is returned otherwise.
* `get/1`: retrieves the value stored with Key in the database.
@ -41,64 +53,60 @@ Usage
```
$ make
$ ./start.sh
%% create a new database
1> {ok, Handle} = emdb:open("/tmp/emdb1").
%% create a new database
1> {ok, Handle} = lmdb:open("/tmp/lmdb1").
%% insert the key <<"a">> with value <<"1">>
2> ok = emdb:put(Handle, <<"a">>, <<"1">>).
%% insert the key <<"a">> with value <<"1">>
2> ok = lmdb:put(Handle, <<"a">>, <<"1">>).
%% try to re-insert the same key <<"a">>
3> key_exist = emdb:put(Handle, <<"a">>, <<"2">>).
%% try to re-insert the same key <<"a">>
3> key_exist = lmdb:put(Handle, <<"a">>, <<"2">>).
%% add a new key-value pair
4> ok = emdb:put(Handle, <<"b">>, <<"2">>).
%% add a new key-value pair
4> ok = lmdb:put(Handle, <<"b">>, <<"2">>).
%% search a non-existing key <<"c">>
5> none = emdb:get(Handle, <<"c">>).
%% search a non-existing key <<"c">>
5> none = lmdb:get(Handle, <<"c">>).
%% retrieve the value for key <<"b">>
6> {ok, <<"2">>} = emdb:get(Handle, <<"b">>).
%% retrieve the value for key <<"b">>
6> {ok, <<"2">>} = lmdb:get(Handle, <<"b">>).
%% retrieve the value for key <<"a">>
7> {ok, <<"1">>} = emdb:get(Handle, <<"a">>).
%% retrieve the value for key <<"a">>
7> {ok, <<"1">>} = lmdb:get(Handle, <<"a">>).
%% delete key <<"b">>
8> ok = emdb:del(Handle, <<"b">>).
%% delete key <<"b">>
8> ok = lmdb:del(Handle, <<"b">>).
%% search a non-existing key <<"b">>
9> none = emdb:get(Handle, <<"b">>).
%% search a non-existing key <<"b">>
9> none = lmdb:get(Handle, <<"b">>).
%% delete a non-existing key <<"z">>
10> none = emdb:del(Handle, <<"z">>).
%% delete a non-existing key <<"z">>
10> none = lmdb:del(Handle, <<"z">>).
%% ensure key <<"a">>'s value is still <<"1">>
11> {ok, <<"1">>} = emdb:get(Handle, <<"a">>).
%% ensure key <<"a">>'s value is still <<"1">>
11> {ok, <<"1">>} = lmdb:get(Handle, <<"a">>).
%% update the value for key <<"a">>
12> ok = emdb:update(Handle, <<"a">>, <<"7">>).
%% update the value for key <<"a">>
12> ok = lmdb:update(Handle, <<"a">>, <<"7">>).
%% check the new value for key <<"a">>
13> {ok, <<"7">>} = emdb:get(Handle, <<"a">>).
%% check the new value for key <<"a">>
13> {ok, <<"7">>} = lmdb:get(Handle, <<"a">>).
%% delete all key-value pairs in the database
14> ok = emdb:drop(Handle).
%% delete all key-value pairs in the database
14> ok = lmdb:drop(Handle).
%% try to retrieve key <<"a">> value
15> none = emdb:get(Handle, <<"a">>).
%% try to retrieve key <<"a">> value
15> none = lmdb:get(Handle, <<"a">>).
%% close the database
16> ok = emdb:close(Handle).
...
17> q().
%% close the database
16> ok = lmdb:close(Handle).
```
#### Note:
The code below creates a new database with **80GB** MapSize, **avoids fsync** after each commit (for an "ACI" but not "D" database we trade durability for speed) and uses the experimental **MDB_FIXEDMAP**.
```
{ok, Handle} = emdb:open("/tmp/emdb2", 85899345920, ?MDB_NOSYNC bor ?MDB_FIXEDMAP).
{ok, Handle} = lmdb:open("/tmp/lmdb2", 85899345920, ?MDB_NOSYNC bor ?MDB_FIXEDMAP).
```
Performance
@ -109,9 +117,10 @@ See the [microbench](http://highlandsun.com/hyc/mdb/microbench/) against:
* SQLite3
* Kyoto TreeDB
* BerkeleyDB 5.x
* btree2n.c (?)
* WiredTiger (?)
MDB performs mmap's the database, so unless your dataset is < 2^32 bytes you'll
need to run on a 64-bit arch system.
MDB will mmap the entire database into memory which means that if your dataset is larger than 2^32 bytes you'll need to run on a 64-bit arch system.
Supported Operating Systems
@ -120,8 +129,8 @@ Supported Operating Systems
Should work on:
* Linux
* OSX
* FreeBSD
* Mac OS/X
* *BSD
* Windows
TODO
@ -134,25 +143,26 @@ TODO
* basho_bench driver
* EQC, PULSE testing
* Key expirey
* renaming
* emdb -> lmdb
* emdb.c -> lmdb_nif.c
* improve stats
* txn API
* cursor API
* config
* use async_nif affinity
* riak_kv backend
* use dups
* 2i
* aae alternative
Other Ideas
-----------
* Create a Riak/KV [backend](http://wiki.basho.com/Storage-Backends.html)
* use dups for siblings
* use txn for 2i
* Create a Riak/KV AAE (riak_kv/src/hashtree.erl) alternative using LMDB
Status
------
Work in progress, not production quality and not supported by Basho Technologies.
Work in progress, not production quality and not supported by Basho Technologies. This is an experiment at this time, nothing more. You are encouraged to contribute code, tests, etc. as you see fit.
LICENSE
-------
EMDB is Copyright (C) 2012-2013 by Aleph Archives and Basho Technologies, Inc., and released under the [OpenLDAP](http://www.OpenLDAP.org/license.html) License.
LMDB is Copyright (C) 2012 by Aleph Archives and (C) 2013 by Basho Technologies, Inc., and released under the [OpenLDAP](http://www.OpenLDAP.org/license.html) License.

155
c_src/emdb.c → c_src/lmdb_nif.c
View file

@ -1,7 +1,8 @@
/* -------------------------------------------------------------------------
* This file is part of EMDB - Erlang MDB API
* This file is part of LMDB - Erlang Lightning MDB API
*
* Copyright (c) 2012 by Aleph Archives. All rights reserved.
%% Copyright (c) 2013 by Basho Technologies, Inc. All rights reserved.
*
* -------------------------------------------------------------------------
* Redistribution and use in source and binary forms, with or without
@ -37,27 +38,27 @@
#include "stats.h"
#include "lmdb.h"
STAT_DECL(emdb_get, 1000);
STAT_DECL(emdb_put, 1000);
STAT_DECL(emdb_del, 1000);
STAT_DECL(emdb_upd, 1000);
STAT_DECL(lmdb_get, 1000);
STAT_DECL(lmdb_put, 1000);
STAT_DECL(lmdb_del, 1000);
STAT_DECL(lmdb_upd, 1000);
static ErlNifResourceType *emdb_RESOURCE;
struct emdb {
static ErlNifResourceType *lmdb_RESOURCE;
struct lmdb {
MDB_env *env;
MDB_dbi dbi;
STAT_DEF(emdb_get);
STAT_DEF(emdb_put);
STAT_DEF(emdb_del);
STAT_DEF(emdb_upd);
STAT_DEF(lmdb_get);
STAT_DEF(lmdb_put);
STAT_DEF(lmdb_del);
STAT_DEF(lmdb_upd);
};
struct emdb_priv_data {
struct lmdb_priv_data {
void *async_nif_priv; // Note: must be first element in struct
};
/* Global init for async_nif. */
ASYNC_NIF_INIT(emdb);
ASYNC_NIF_INIT(lmdb);
/* Atoms (initialized in on_load) */
static ERL_NIF_TERM ATOM_ERROR;
@ -182,7 +183,7 @@ __strerror_term(ErlNifEnv* env, int err)
* argv[2] flags
*/
ASYNC_NIF_DECL(
emdb_open,
lmdb_open,
{ // struct
char dirname[MAXPATHLEN];
@ -206,16 +207,16 @@ ASYNC_NIF_DECL(
ERL_NIF_TERM err;
MDB_txn *txn;
struct emdb *handle;
struct lmdb *handle;
int ret;
if ((handle = enif_alloc_resource(emdb_RESOURCE, sizeof(struct emdb))) == NULL)
if ((handle = enif_alloc_resource(lmdb_RESOURCE, sizeof(struct lmdb))) == NULL)
FAIL_ERR(ENOMEM, err3);
STAT_INIT(handle, emdb_get);
STAT_INIT(handle, emdb_put);
STAT_INIT(handle, emdb_upd);
STAT_INIT(handle, emdb_del);
STAT_INIT(handle, lmdb_get);
STAT_INIT(handle, lmdb_put);
STAT_INIT(handle, lmdb_upd);
STAT_INIT(handle, lmdb_del);
CHECK(mdb_env_create(&(handle->env)), err2);
@ -253,15 +254,15 @@ ASYNC_NIF_DECL(
* argv[0] reference to the MDB handle resource
*/
ASYNC_NIF_DECL(
emdb_close,
lmdb_close,
{ // struct
struct emdb *handle;
struct lmdb *handle;
},
{ // pre
if (!(argc == 1 &&
enif_get_resource(env, argv[0], emdb_RESOURCE, (void**)&args->handle))) {
enif_get_resource(env, argv[0], lmdb_RESOURCE, (void**)&args->handle))) {
ASYNC_NIF_RETURN_BADARG();
}
if (!args->handle->env)
@ -270,15 +271,15 @@ ASYNC_NIF_DECL(
},
{ // work
STAT_PRINT(args->handle, emdb_get, "emdb");
STAT_PRINT(args->handle, emdb_put, "emdb");
STAT_PRINT(args->handle, emdb_del, "emdb");
STAT_PRINT(args->handle, emdb_upd, "emdb");
STAT_PRINT(args->handle, lmdb_get, "lmdb");
STAT_PRINT(args->handle, lmdb_put, "lmdb");
STAT_PRINT(args->handle, lmdb_del, "lmdb");
STAT_PRINT(args->handle, lmdb_upd, "lmdb");
mdb_env_close(args->handle->env);
STAT_RESET(args->handle, emdb_get);
STAT_RESET(args->handle, emdb_put);
STAT_RESET(args->handle, emdb_del);
STAT_RESET(args->handle, emdb_upd);
STAT_RESET(args->handle, lmdb_get);
STAT_RESET(args->handle, lmdb_put);
STAT_RESET(args->handle, lmdb_del);
STAT_RESET(args->handle, lmdb_upd);
args->handle->env = NULL;
ASYNC_NIF_REPLY(ATOM_OK);
return;
@ -297,24 +298,24 @@ ASYNC_NIF_DECL(
* argv[2] value as an Erlang binary
*/
ASYNC_NIF_DECL(
emdb_put,
lmdb_put,
{ // struct
struct emdb *handle;
struct lmdb *handle;
ERL_NIF_TERM key;
ERL_NIF_TERM val;
},
{ // pre
if (!(argc == 3 &&
enif_get_resource(env, argv[0], emdb_RESOURCE, (void**)&args->handle) &&
enif_get_resource(env, argv[0], lmdb_RESOURCE, (void**)&args->handle) &&
enif_is_binary(env, argv[1]) &&
enif_is_binary(env, argv[2]) )) {
ASYNC_NIF_RETURN_BADARG();
}
if (!args->handle->env)
ASYNC_NIF_RETURN_BADARG();
STAT_TICK(args->handle, emdb_put);
STAT_TICK(args->handle, lmdb_put);
enif_keep_resource((void*)args->handle);
args->key = enif_make_copy(ASYNC_NIF_WORK_ENV, argv[1]);
args->val = enif_make_copy(ASYNC_NIF_WORK_ENV, argv[2]);
@ -353,7 +354,7 @@ ASYNC_NIF_DECL(
FAIL_ERR(ret, err1);
CHECK(mdb_txn_commit(txn), err1);
STAT_TOCK(args->handle, emdb_put);
STAT_TOCK(args->handle, lmdb_put);
ASYNC_NIF_REPLY(ATOM_OK);
return;
@ -377,24 +378,24 @@ ASYNC_NIF_DECL(
* argv[2] value as an Erlang binary
*/
ASYNC_NIF_DECL(
emdb_update,
lmdb_update,
{ // struct
struct emdb *handle;
struct lmdb *handle;
ERL_NIF_TERM key;
ERL_NIF_TERM val;
},
{ // pre
if (!(argc == 3 &&
enif_get_resource(env, argv[0], emdb_RESOURCE, (void**)&args->handle) &&
enif_get_resource(env, argv[0], lmdb_RESOURCE, (void**)&args->handle) &&
enif_is_binary(env, argv[1]) &&
enif_is_binary(env, argv[2]) )) {
ASYNC_NIF_RETURN_BADARG();
}
if (!args->handle->env)
ASYNC_NIF_RETURN_BADARG();
STAT_TICK(args->handle, emdb_upd);
STAT_TICK(args->handle, lmdb_upd);
enif_keep_resource((void*)args->handle);
args->key = enif_make_copy(ASYNC_NIF_WORK_ENV, argv[1]);
args->val = enif_make_copy(ASYNC_NIF_WORK_ENV, argv[2]);
@ -426,7 +427,7 @@ ASYNC_NIF_DECL(
CHECK(mdb_txn_begin(args->handle->env, NULL, 0, & txn), err2);
CHECK(mdb_put(txn, args->handle->dbi, &mkey, &mdata, 0), err1);
CHECK(mdb_txn_commit(txn), err1);
STAT_TOCK(args->handle, emdb_upd);
STAT_TOCK(args->handle, lmdb_upd);
ASYNC_NIF_REPLY(ATOM_OK);
return;
@ -449,22 +450,22 @@ ASYNC_NIF_DECL(
* argv[1] key as an Erlang binary
*/
ASYNC_NIF_DECL(
emdb_get,
lmdb_get,
{ // struct
struct emdb *handle;
struct lmdb *handle;
ERL_NIF_TERM key;
},
{ // pre
if (!(argc == 2 &&
enif_get_resource(env, argv[0], emdb_RESOURCE, (void**)&args->handle) &&
enif_get_resource(env, argv[0], lmdb_RESOURCE, (void**)&args->handle) &&
enif_is_binary(env, argv[1]) )) {
ASYNC_NIF_RETURN_BADARG();
}
if (!args->handle->env)
ASYNC_NIF_RETURN_BADARG();
STAT_TICK(args->handle, emdb_get);
STAT_TICK(args->handle, lmdb_get);
enif_keep_resource((void*)args->handle);
args->key = enif_make_copy(ASYNC_NIF_WORK_ENV, argv[1]);
},
@ -504,7 +505,7 @@ ASYNC_NIF_DECL(
FAIL_ERR(ENOMEM, err);
memcpy(bin, mdata.mv_data, mdata.mv_size);
STAT_TOCK(args->handle, emdb_get);
STAT_TOCK(args->handle, lmdb_get);
ASYNC_NIF_REPLY(enif_make_tuple(env, 2, ATOM_OK, val));
return;
@ -525,22 +526,22 @@ ASYNC_NIF_DECL(
* argv[1] key as an Erlang binary
*/
ASYNC_NIF_DECL(
emdb_del,
lmdb_del,
{ // struct
struct emdb *handle;
struct lmdb *handle;
ERL_NIF_TERM key;
},
{ // pre
if (!(argc == 2 &&
enif_get_resource(env, argv[0], emdb_RESOURCE, (void**)&args->handle) &&
enif_get_resource(env, argv[0], lmdb_RESOURCE, (void**)&args->handle) &&
enif_is_binary(env, argv[1]) )) {
ASYNC_NIF_RETURN_BADARG();
}
if (!args->handle->env)
ASYNC_NIF_RETURN_BADARG();
STAT_TICK(args->handle, emdb_del);
STAT_TICK(args->handle, lmdb_del);
enif_keep_resource((void*)args->handle);
args->key = enif_make_copy(ASYNC_NIF_WORK_ENV, argv[1]);
},
@ -570,7 +571,7 @@ ASYNC_NIF_DECL(
}
CHECK(mdb_txn_commit(txn), err);
STAT_TOCK(args->handle, emdb_del);
STAT_TOCK(args->handle, lmdb_del);
ASYNC_NIF_REPLY(ATOM_OK);
return;
@ -590,15 +591,15 @@ ASYNC_NIF_DECL(
* argv[0] reference to the MDB handle resource
*/
ASYNC_NIF_DECL(
emdb_drop,
lmdb_drop,
{ // struct
struct emdb *handle;
struct lmdb *handle;
},
{ // pre
if (!(argc == 1 &&
enif_get_resource(env, argv[0], emdb_RESOURCE, (void**)&args->handle))) {
enif_get_resource(env, argv[0], lmdb_RESOURCE, (void**)&args->handle))) {
ASYNC_NIF_RETURN_BADARG();
}
if (!args->handle->env)
@ -631,20 +632,20 @@ ASYNC_NIF_DECL(
static int emdb_load(ErlNifEnv* env, void** priv_data, ERL_NIF_TERM load_info)
static int lmdb_load(ErlNifEnv* env, void** priv_data, ERL_NIF_TERM load_info)
{
__UNUSED(load_info);
ErlNifResourceFlags flags = ERL_NIF_RT_CREATE | ERL_NIF_RT_TAKEOVER;
struct emdb_priv_data *priv = enif_alloc(sizeof(struct emdb_priv_data));
struct lmdb_priv_data *priv = enif_alloc(sizeof(struct lmdb_priv_data));
if (!priv)
return ENOMEM;
memset(priv, 0, sizeof(struct emdb_priv_data));
memset(priv, 0, sizeof(struct lmdb_priv_data));
/* Note: !!! the first element of our priv_data struct *must* be the
pointer to the async_nif's private data which we set here. */
ASYNC_NIF_LOAD(emdb, priv->async_nif_priv);
ASYNC_NIF_LOAD(lmdb, priv->async_nif_priv);
if (!priv)
return ENOMEM;
*priv_data = priv;
@ -670,12 +671,12 @@ static int emdb_load(ErlNifEnv* env, void** priv_data, ERL_NIF_TERM load_info)
ATOM_INCOMPATIBLE = enif_make_atom(env, "incompatible");
ATOM_BAD_RSLOT = enif_make_atom(env, "bad_rslot");
emdb_RESOURCE = enif_open_resource_type(env, NULL, "emdb_resource",
lmdb_RESOURCE = enif_open_resource_type(env, NULL, "lmdb_resource",
NULL, flags, NULL);
return (0);
}
static int emdb_reload(ErlNifEnv* env, void** priv_data, ERL_NIF_TERM info)
static int lmdb_reload(ErlNifEnv* env, void** priv_data, ERL_NIF_TERM info)
{
__UNUSED(env);
__UNUSED(priv_data);
@ -684,39 +685,39 @@ static int emdb_reload(ErlNifEnv* env, void** priv_data, ERL_NIF_TERM info)
}
static int emdb_upgrade(ErlNifEnv* env, void** priv_data, void** old_priv, ERL_NIF_TERM load_info)
static int lmdb_upgrade(ErlNifEnv* env, void** priv_data, void** old_priv, ERL_NIF_TERM load_info)
{
__UNUSED(env);
__UNUSED(priv_data);
__UNUSED(old_priv);
__UNUSED(load_info);
ASYNC_NIF_UPGRADE(emdb, env);
ASYNC_NIF_UPGRADE(lmdb, env);
return (0); // TODO:
}
static void emdb_unload(ErlNifEnv* env, void* priv_data)
static void lmdb_unload(ErlNifEnv* env, void* priv_data)
{
struct emdb_priv_data *priv = (struct emdb_priv_data *)priv_data;
ASYNC_NIF_UNLOAD(emdb, env, priv->async_nif_priv);
struct lmdb_priv_data *priv = (struct lmdb_priv_data *)priv_data;
ASYNC_NIF_UNLOAD(lmdb, env, priv->async_nif_priv);
enif_free(priv);
return;
}
static ErlNifFunc nif_funcs [] = {
{"open", 4, emdb_open},
{"close", 2, emdb_close},
{"put", 4, emdb_put},
{"get", 3, emdb_get},
{"del", 3, emdb_del},
{"update", 4, emdb_update},
{"drop", 2, emdb_drop}
{"open", 4, lmdb_open},
{"close", 2, lmdb_close},
{"put", 4, lmdb_put},
{"get", 3, lmdb_get},
{"del", 3, lmdb_del},
{"update", 4, lmdb_update},
{"drop", 2, lmdb_drop}
};
/* driver entry point */
ERL_NIF_INIT(emdb,
ERL_NIF_INIT(lmdb,
nif_funcs,
& emdb_load,
& emdb_reload,
& emdb_upgrade,
& emdb_unload)
& lmdb_load,
& lmdb_reload,
& lmdb_upgrade,
& lmdb_unload)

10
c_src/stats.h
View file

@ -128,9 +128,13 @@ static unsigned int __log2_64(uint64_t x) {
{ \
uint8_t logs[64]; \
uint8_t i, j, max_log = 0; \
double m = (s->mean + name ## _stat_mean(s) / 2.0); \
double m = 0.0; \
\
fprintf(stderr, "%s:async_nif request latency histogram:\n", mod); \
if (s->n < nsamples) \
return; \
\
fprintf(stderr, "\n%s:async_nif request latency histogram:\n", mod); \
m = (s->mean + name ## _stat_mean(s) / 2.0); \
for (i = 0; i < 64; i++) { \
logs[i] = LOG2(s->histogram[i]); \
if (logs[i] > max_log) \
@ -145,7 +149,7 @@ static unsigned int __log2_64(uint64_t x) {
fprintf(stderr, logs[j] >= i ? "•" : " "); \
fprintf(stderr, "\n"); \
} \
if (max_log == 0) { \
if (max_log == 100) { \
fprintf(stderr, "[empty]\n"); \
} else { \
fprintf(stderr, " ns μs ms s ks\n"); \

915
c_src/uthash.h
View file

@ -1,915 +0,0 @@
/*
Copyright (c) 2003-2012, Troy D. Hanson http://uthash.sourceforge.net
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS
IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#ifndef UTHASH_H
#define UTHASH_H
#include <string.h> /* memcmp,strlen */
#include <stddef.h> /* ptrdiff_t */
#include <stdlib.h> /* exit() */
/* These macros use decltype or the earlier __typeof GNU extension.
As decltype is only available in newer compilers (VS2010 or gcc 4.3+
when compiling c++ source) this code uses whatever method is needed
or, for VS2008 where neither is available, uses casting workarounds. */
#ifdef _MSC_VER /* MS compiler */
#if _MSC_VER >= 1600 && defined(__cplusplus) /* VS2010 or newer in C++ mode */
#define DECLTYPE(x) (decltype(x))
#else /* VS2008 or older (or VS2010 in C mode) */
#define NO_DECLTYPE
#define DECLTYPE(x)
#endif
#else /* GNU, Sun and other compilers */
#define DECLTYPE(x) (__typeof(x))
#endif
#ifdef NO_DECLTYPE
#define DECLTYPE_ASSIGN(dst,src) \
do { \
char **_da_dst = (char**)(&(dst)); \
*_da_dst = (char*)(src); \
} while(0)
#else
#define DECLTYPE_ASSIGN(dst,src) \
do { \
(dst) = DECLTYPE(dst)(src); \
} while(0)
#endif
/* a number of the hash function use uint32_t which isn't defined on win32 */
#ifdef _MSC_VER
typedef unsigned int uint32_t;
typedef unsigned char uint8_t;
#else
#include <inttypes.h> /* uint32_t */
#endif
#define UTHASH_VERSION 1.9.6
#ifndef uthash_fatal
#define uthash_fatal(msg) exit(-1) /* fatal error (out of memory,etc) */
#endif
#ifndef uthash_malloc
#define uthash_malloc(sz) malloc(sz) /* malloc fcn */
#endif
#ifndef uthash_free
#define uthash_free(ptr,sz) free(ptr) /* free fcn */
#endif
#ifndef uthash_noexpand_fyi
#define uthash_noexpand_fyi(tbl) /* can be defined to log noexpand */
#endif
#ifndef uthash_expand_fyi
#define uthash_expand_fyi(tbl) /* can be defined to log expands */
#endif
/* initial number of buckets */
#define HASH_INITIAL_NUM_BUCKETS 32 /* initial number of buckets */
#define HASH_INITIAL_NUM_BUCKETS_LOG2 5 /* lg2 of initial number of buckets */
#define HASH_BKT_CAPACITY_THRESH 10 /* expand when bucket count reaches */
/* calculate the element whose hash handle address is hhe */
#define ELMT_FROM_HH(tbl,hhp) ((void*)(((char*)(hhp)) - ((tbl)->hho)))
#define HASH_FIND(hh,head,keyptr,keylen,out) \
do { \
unsigned _hf_bkt,_hf_hashv; \
out=NULL; \
if (head) { \
HASH_FCN(keyptr,keylen, (head)->hh.tbl->num_buckets, _hf_hashv, _hf_bkt); \
if (HASH_BLOOM_TEST((head)->hh.tbl, _hf_hashv)) { \
HASH_FIND_IN_BKT((head)->hh.tbl, hh, (head)->hh.tbl->buckets[ _hf_bkt ], \
keyptr,keylen,out); \
} \
} \
} while (0)
#ifdef HASH_BLOOM
#define HASH_BLOOM_BITLEN (1ULL << HASH_BLOOM)
#define HASH_BLOOM_BYTELEN (HASH_BLOOM_BITLEN/8) + ((HASH_BLOOM_BITLEN%8) ? 1:0)
#define HASH_BLOOM_MAKE(tbl) \
do { \
(tbl)->bloom_nbits = HASH_BLOOM; \
(tbl)->bloom_bv = (uint8_t*)uthash_malloc(HASH_BLOOM_BYTELEN); \
if (!((tbl)->bloom_bv)) { uthash_fatal( "out of memory"); } \
memset((tbl)->bloom_bv, 0, HASH_BLOOM_BYTELEN); \
(tbl)->bloom_sig = HASH_BLOOM_SIGNATURE; \
} while (0)
#define HASH_BLOOM_FREE(tbl) \
do { \
uthash_free((tbl)->bloom_bv, HASH_BLOOM_BYTELEN); \
} while (0)
#define HASH_BLOOM_BITSET(bv,idx) (bv[(idx)/8] |= (1U << ((idx)%8)))
#define HASH_BLOOM_BITTEST(bv,idx) (bv[(idx)/8] & (1U << ((idx)%8)))
#define HASH_BLOOM_ADD(tbl,hashv) \
HASH_BLOOM_BITSET((tbl)->bloom_bv, (hashv & (uint32_t)((1ULL << (tbl)->bloom_nbits) - 1)))
#define HASH_BLOOM_TEST(tbl,hashv) \
HASH_BLOOM_BITTEST((tbl)->bloom_bv, (hashv & (uint32_t)((1ULL << (tbl)->bloom_nbits) - 1)))
#else
#define HASH_BLOOM_MAKE(tbl)
#define HASH_BLOOM_FREE(tbl)
#define HASH_BLOOM_ADD(tbl,hashv)
#define HASH_BLOOM_TEST(tbl,hashv) (1)
#endif
#define HASH_MAKE_TABLE(hh,head) \
do { \
(head)->hh.tbl = (UT_hash_table*)uthash_malloc( \
sizeof(UT_hash_table)); \
if (!((head)->hh.tbl)) { uthash_fatal( "out of memory"); } \
memset((head)->hh.tbl, 0, sizeof(UT_hash_table)); \
(head)->hh.tbl->tail = &((head)->hh); \
(head)->hh.tbl->num_buckets = HASH_INITIAL_NUM_BUCKETS; \
(head)->hh.tbl->log2_num_buckets = HASH_INITIAL_NUM_BUCKETS_LOG2; \
(head)->hh.tbl->hho = (char*)(&(head)->hh) - (char*)(head); \
(head)->hh.tbl->buckets = (UT_hash_bucket*)uthash_malloc( \
HASH_INITIAL_NUM_BUCKETS*sizeof(struct UT_hash_bucket)); \
if (! (head)->hh.tbl->buckets) { uthash_fatal( "out of memory"); } \
memset((head)->hh.tbl->buckets, 0, \
HASH_INITIAL_NUM_BUCKETS*sizeof(struct UT_hash_bucket)); \
HASH_BLOOM_MAKE((head)->hh.tbl); \
(head)->hh.tbl->signature = HASH_SIGNATURE; \
} while(0)
#define HASH_ADD(hh,head,fieldname,keylen_in,add) \
HASH_ADD_KEYPTR(hh,head,&((add)->fieldname),keylen_in,add)
#define HASH_ADD_KEYPTR(hh,head,keyptr,keylen_in,add) \
do { \
unsigned _ha_bkt; \
(add)->hh.next = NULL; \
(add)->hh.key = (char*)keyptr; \
(add)->hh.keylen = (unsigned)keylen_in; \
if (!(head)) { \
head = (add); \
(head)->hh.prev = NULL; \
HASH_MAKE_TABLE(hh,head); \
} else { \
(head)->hh.tbl->tail->next = (add); \
(add)->hh.prev = ELMT_FROM_HH((head)->hh.tbl, (head)->hh.tbl->tail); \
(head)->hh.tbl->tail = &((add)->hh); \
} \
(head)->hh.tbl->num_items++; \
(add)->hh.tbl = (head)->hh.tbl; \
HASH_FCN(keyptr,keylen_in, (head)->hh.tbl->num_buckets, \
(add)->hh.hashv, _ha_bkt); \
HASH_ADD_TO_BKT((head)->hh.tbl->buckets[_ha_bkt],&(add)->hh); \
HASH_BLOOM_ADD((head)->hh.tbl,(add)->hh.hashv); \
HASH_EMIT_KEY(hh,head,keyptr,keylen_in); \
HASH_FSCK(hh,head); \
} while(0)
#define HASH_TO_BKT( hashv, num_bkts, bkt ) \
do { \
bkt = ((hashv) & ((num_bkts) - 1)); \
} while(0)
/* delete "delptr" from the hash table.
* "the usual" patch-up process for the app-order doubly-linked-list.
* The use of _hd_hh_del below deserves special explanation.
* These used to be expressed using (delptr) but that led to a bug
* if someone used the same symbol for the head and deletee, like
* HASH_DELETE(hh,users,users);
* We want that to work, but by changing the head (users) below
* we were forfeiting our ability to further refer to the deletee (users)
* in the patch-up process. Solution: use scratch space to
* copy the deletee pointer, then the latter references are via that
* scratch pointer rather than through the repointed (users) symbol.
*/
#define HASH_DELETE(hh,head,delptr) \
do { \
unsigned _hd_bkt; \
struct UT_hash_handle *_hd_hh_del; \
if ( ((delptr)->hh.prev == NULL) && ((delptr)->hh.next == NULL) ) { \
uthash_free((head)->hh.tbl->buckets, \
(head)->hh.tbl->num_buckets*sizeof(struct UT_hash_bucket) ); \
HASH_BLOOM_FREE((head)->hh.tbl); \
uthash_free((head)->hh.tbl, sizeof(UT_hash_table)); \
head = NULL; \
} else { \
_hd_hh_del = &((delptr)->hh); \
if ((delptr) == ELMT_FROM_HH((head)->hh.tbl,(head)->hh.tbl->tail)) { \
(head)->hh.tbl->tail = \
(UT_hash_handle*)((char*)((delptr)->hh.prev) + \
(head)->hh.tbl->hho); \
} \
if ((delptr)->hh.prev) { \
((UT_hash_handle*)((char*)((delptr)->hh.prev) + \
(head)->hh.tbl->hho))->next = (delptr)->hh.next; \
} else { \
DECLTYPE_ASSIGN(head,(delptr)->hh.next); \
} \
if (_hd_hh_del->next) { \
((UT_hash_handle*)((char*)_hd_hh_del->next + \
(head)->hh.tbl->hho))->prev = \
_hd_hh_del->prev; \
} \
HASH_TO_BKT( _hd_hh_del->hashv, (head)->hh.tbl->num_buckets, _hd_bkt); \
HASH_DEL_IN_BKT(hh,(head)->hh.tbl->buckets[_hd_bkt], _hd_hh_del); \
(head)->hh.tbl->num_items--; \
} \
HASH_FSCK(hh,head); \
} while (0)
/* convenience forms of HASH_FIND/HASH_ADD/HASH_DEL */
#define HASH_FIND_STR(head,findstr,out) \
HASH_FIND(hh,head,findstr,strlen(findstr),out)
#define HASH_ADD_STR(head,strfield,add) \
HASH_ADD(hh,head,strfield,strlen(add->strfield),add)
#define HASH_FIND_INT(head,findint,out) \
HASH_FIND(hh,head,findint,sizeof(int),out)
#define HASH_ADD_INT(head,intfield,add) \
HASH_ADD(hh,head,intfield,sizeof(int),add)
#define HASH_FIND_PTR(head,findptr,out) \
HASH_FIND(hh,head,findptr,sizeof(void *),out)
#define HASH_ADD_PTR(head,ptrfield,add) \
HASH_ADD(hh,head,ptrfield,sizeof(void *),add)
#define HASH_DEL(head,delptr) \
HASH_DELETE(hh,head,delptr)
/* HASH_FSCK checks hash integrity on every add/delete when HASH_DEBUG is defined.
* This is for uthash developer only; it compiles away if HASH_DEBUG isn't defined.
*/
#ifdef HASH_DEBUG
#define HASH_OOPS(...) do { fprintf(stderr,__VA_ARGS__); exit(-1); } while (0)
#define HASH_FSCK(hh,head) \
do { \
unsigned _bkt_i; \
unsigned _count, _bkt_count; \
char *_prev; \
struct UT_hash_handle *_thh; \
if (head) { \
_count = 0; \
for( _bkt_i = 0; _bkt_i < (head)->hh.tbl->num_buckets; _bkt_i++) { \
_bkt_count = 0; \
_thh = (head)->hh.tbl->buckets[_bkt_i].hh_head; \
_prev = NULL; \
while (_thh) { \
if (_prev != (char*)(_thh->hh_prev)) { \
HASH_OOPS("invalid hh_prev %p, actual %p\n", \
_thh->hh_prev, _prev ); \
} \
_bkt_count++; \
_prev = (char*)(_thh); \
_thh = _thh->hh_next; \
} \
_count += _bkt_count; \
if ((head)->hh.tbl->buckets[_bkt_i].count != _bkt_count) { \
HASH_OOPS("invalid bucket count %d, actual %d\n", \
(head)->hh.tbl->buckets[_bkt_i].count, _bkt_count); \
} \
} \
if (_count != (head)->hh.tbl->num_items) { \
HASH_OOPS("invalid hh item count %d, actual %d\n", \
(head)->hh.tbl->num_items, _count ); \
} \
/* traverse hh in app order; check next/prev integrity, count */ \
_count = 0; \
_prev = NULL; \
_thh = &(head)->hh; \
while (_thh) { \
_count++; \
if (_prev !=(char*)(_thh->prev)) { \
HASH_OOPS("invalid prev %p, actual %p\n", \
_thh->prev, _prev ); \
} \
_prev = (char*)ELMT_FROM_HH((head)->hh.tbl, _thh); \
_thh = ( _thh->next ? (UT_hash_handle*)((char*)(_thh->next) + \
(head)->hh.tbl->hho) : NULL ); \
} \
if (_count != (head)->hh.tbl->num_items) { \
HASH_OOPS("invalid app item count %d, actual %d\n", \
(head)->hh.tbl->num_items, _count ); \
} \
} \
} while (0)
#else
#define HASH_FSCK(hh,head)
#endif
/* When compiled with -DHASH_EMIT_KEYS, length-prefixed keys are emitted to
* the descriptor to which this macro is defined for tuning the hash function.
* The app can #include <unistd.h> to get the prototype for write(2). */
#ifdef HASH_EMIT_KEYS
#define HASH_EMIT_KEY(hh,head,keyptr,fieldlen) \
do { \
unsigned _klen = fieldlen; \
write(HASH_EMIT_KEYS, &_klen, sizeof(_klen)); \
write(HASH_EMIT_KEYS, keyptr, fieldlen); \
} while (0)
#else
#define HASH_EMIT_KEY(hh,head,keyptr,fieldlen)
#endif
/* default to Jenkin's hash unless overridden e.g. DHASH_FUNCTION=HASH_SAX */
#ifdef HASH_FUNCTION
#define HASH_FCN HASH_FUNCTION
#else
#define HASH_FCN HASH_JEN
#endif
/* The Bernstein hash function, used in Perl prior to v5.6 */
#define HASH_BER(key,keylen,num_bkts,hashv,bkt) \
do { \
unsigned _hb_keylen=keylen; \
char *_hb_key=(char*)(key); \
(hashv) = 0; \
while (_hb_keylen--) { (hashv) = ((hashv) * 33) + *_hb_key++; } \
bkt = (hashv) & (num_bkts-1); \
} while (0)
/* SAX/FNV/OAT/JEN hash functions are macro variants of those listed at
* http://eternallyconfuzzled.com/tuts/algorithms/jsw_tut_hashing.aspx */
#define HASH_SAX(key,keylen,num_bkts,hashv,bkt) \
do { \
unsigned _sx_i; \
char *_hs_key=(char*)(key); \
hashv = 0; \
for(_sx_i=0; _sx_i < keylen; _sx_i++) \
hashv ^= (hashv << 5) + (hashv >> 2) + _hs_key[_sx_i]; \
bkt = hashv & (num_bkts-1); \
} while (0)
#define HASH_FNV(key,keylen,num_bkts,hashv,bkt) \
do { \
unsigned _fn_i; \
char *_hf_key=(char*)(key); \
hashv = 2166136261UL; \
for(_fn_i=0; _fn_i < keylen; _fn_i++) \
hashv = (hashv * 16777619) ^ _hf_key[_fn_i]; \
bkt = hashv & (num_bkts-1); \
} while(0)
#define HASH_OAT(key,keylen,num_bkts,hashv,bkt) \
do { \
unsigned _ho_i; \
char *_ho_key=(char*)(key); \
hashv = 0; \
for(_ho_i=0; _ho_i < keylen; _ho_i++) { \
hashv += _ho_key[_ho_i]; \
hashv += (hashv << 10); \
hashv ^= (hashv >> 6); \
} \
hashv += (hashv << 3); \
hashv ^= (hashv >> 11); \
hashv += (hashv << 15); \
bkt = hashv & (num_bkts-1); \
} while(0)
#define HASH_JEN_MIX(a,b,c) \
do { \
a -= b; a -= c; a ^= ( c >> 13 ); \
b -= c; b -= a; b ^= ( a << 8 ); \
c -= a; c -= b; c ^= ( b >> 13 ); \
a -= b; a -= c; a ^= ( c >> 12 ); \
b -= c; b -= a; b ^= ( a << 16 ); \
c -= a; c -= b; c ^= ( b >> 5 ); \
a -= b; a -= c; a ^= ( c >> 3 ); \
b -= c; b -= a; b ^= ( a << 10 ); \
c -= a; c -= b; c ^= ( b >> 15 ); \
} while (0)
#define HASH_JEN(key,keylen,num_bkts,hashv,bkt) \
do { \
unsigned _hj_i,_hj_j,_hj_k; \
char *_hj_key=(char*)(key); \
hashv = 0xfeedbeef; \
_hj_i = _hj_j = 0x9e3779b9; \
_hj_k = (unsigned)keylen; \
while (_hj_k >= 12) { \
_hj_i += (_hj_key[0] + ( (unsigned)_hj_key[1] << 8 ) \
+ ( (unsigned)_hj_key[2] << 16 ) \
+ ( (unsigned)_hj_key[3] << 24 ) ); \
_hj_j += (_hj_key[4] + ( (unsigned)_hj_key[5] << 8 ) \
+ ( (unsigned)_hj_key[6] << 16 ) \
+ ( (unsigned)_hj_key[7] << 24 ) ); \
hashv += (_hj_key[8] + ( (unsigned)_hj_key[9] << 8 ) \
+ ( (unsigned)_hj_key[10] << 16 ) \
+ ( (unsigned)_hj_key[11] << 24 ) ); \
\
HASH_JEN_MIX(_hj_i, _hj_j, hashv); \
\
_hj_key += 12; \
_hj_k -= 12; \
} \
hashv += keylen; \
switch ( _hj_k ) { \
case 11: hashv += ( (unsigned)_hj_key[10] << 24 ); \
case 10: hashv += ( (unsigned)_hj_key[9] << 16 ); \
case 9: hashv += ( (unsigned)_hj_key[8] << 8 ); \
case 8: _hj_j += ( (unsigned)_hj_key[7] << 24 ); \
case 7: _hj_j += ( (unsigned)_hj_key[6] << 16 ); \
case 6: _hj_j += ( (unsigned)_hj_key[5] << 8 ); \
case 5: _hj_j += _hj_key[4]; \
case 4: _hj_i += ( (unsigned)_hj_key[3] << 24 ); \
case 3: _hj_i += ( (unsigned)_hj_key[2] << 16 ); \
case 2: _hj_i += ( (unsigned)_hj_key[1] << 8 ); \
case 1: _hj_i += _hj_key[0]; \
} \
HASH_JEN_MIX(_hj_i, _hj_j, hashv); \
bkt = hashv & (num_bkts-1); \
} while(0)
/* The Paul Hsieh hash function */
#undef get16bits
#if (defined(__GNUC__) && defined(__i386__)) || defined(__WATCOMC__) \
|| defined(_MSC_VER) || defined (__BORLANDC__) || defined (__TURBOC__)
#define get16bits(d) (*((const uint16_t *) (d)))
#endif
#if !defined (get16bits)
#define get16bits(d) ((((uint32_t)(((const uint8_t *)(d))[1])) << 8) \
+(uint32_t)(((const uint8_t *)(d))[0]) )
#endif
#define HASH_SFH(key,keylen,num_bkts,hashv,bkt) \
do { \
char *_sfh_key=(char*)(key); \
uint32_t _sfh_tmp, _sfh_len = keylen; \
\
int _sfh_rem = _sfh_len & 3; \
_sfh_len >>= 2; \
hashv = 0xcafebabe; \
\
/* Main loop */ \
for (;_sfh_len > 0; _sfh_len--) { \
hashv += get16bits (_sfh_key); \
_sfh_tmp = (get16bits (_sfh_key+2) << 11) ^ hashv; \
hashv = (hashv << 16) ^ _sfh_tmp; \
_sfh_key += 2*sizeof (uint16_t); \
hashv += hashv >> 11; \
} \
\
/* Handle end cases */ \
switch (_sfh_rem) { \
case 3: hashv += get16bits (_sfh_key); \
hashv ^= hashv << 16; \
hashv ^= _sfh_key[sizeof (uint16_t)] << 18; \
hashv += hashv >> 11; \
break; \
case 2: hashv += get16bits (_sfh_key); \
hashv ^= hashv << 11; \
hashv += hashv >> 17; \
break; \
case 1: hashv += *_sfh_key; \
hashv ^= hashv << 10; \
hashv += hashv >> 1; \
} \
\
/* Force "avalanching" of final 127 bits */ \
hashv ^= hashv << 3; \
hashv += hashv >> 5; \
hashv ^= hashv << 4; \
hashv += hashv >> 17; \
hashv ^= hashv << 25; \
hashv += hashv >> 6; \
bkt = hashv & (num_bkts-1); \
} while(0)
#ifdef HASH_USING_NO_STRICT_ALIASING
/* The MurmurHash exploits some CPU's (x86,x86_64) tolerance for unaligned reads.
* For other types of CPU's (e.g. Sparc) an unaligned read causes a bus error.
* MurmurHash uses the faster approach only on CPU's where we know it's safe.
*
* Note the preprocessor built-in defines can be emitted using:
*
* gcc -m64 -dM -E - < /dev/null (on gcc)
* cc -## a.c (where a.c is a simple test file) (Sun Studio)
*/
#if (defined(__i386__) || defined(__x86_64__))
#define MUR_GETBLOCK(p,i) p[i]
#else /* non intel */
#define MUR_PLUS0_ALIGNED(p) (((unsigned long)p & 0x3) == 0)
#define MUR_PLUS1_ALIGNED(p) (((unsigned long)p & 0x3) == 1)
#define MUR_PLUS2_ALIGNED(p) (((unsigned long)p & 0x3) == 2)
#define MUR_PLUS3_ALIGNED(p) (((unsigned long)p & 0x3) == 3)
#define WP(p) ((uint32_t*)((unsigned long)(p) & ~3UL))
#if (defined(__BIG_ENDIAN__) || defined(SPARC) || defined(__ppc__) || defined(__ppc64__))
#define MUR_THREE_ONE(p) ((((*WP(p))&0x00ffffff) << 8) | (((*(WP(p)+1))&0xff000000) >> 24))
#define MUR_TWO_TWO(p) ((((*WP(p))&0x0000ffff) <<16) | (((*(WP(p)+1))&0xffff0000) >> 16))
#define MUR_ONE_THREE(p) ((((*WP(p))&0x000000ff) <<24) | (((*(WP(p)+1))&0xffffff00) >> 8))
#else /* assume little endian non-intel */
#define MUR_THREE_ONE(p) ((((*WP(p))&0xffffff00) >> 8) | (((*(WP(p)+1))&0x000000ff) << 24))
#define MUR_TWO_TWO(p) ((((*WP(p))&0xffff0000) >>16) | (((*(WP(p)+1))&0x0000ffff) << 16))
#define MUR_ONE_THREE(p) ((((*WP(p))&0xff000000) >>24) | (((*(WP(p)+1))&0x00ffffff) << 8))
#endif
#define MUR_GETBLOCK(p,i) (MUR_PLUS0_ALIGNED(p) ? ((p)[i]) : \
(MUR_PLUS1_ALIGNED(p) ? MUR_THREE_ONE(p) : \
(MUR_PLUS2_ALIGNED(p) ? MUR_TWO_TWO(p) : \
MUR_ONE_THREE(p))))
#endif
#define MUR_ROTL32(x,r) (((x) << (r)) | ((x) >> (32 - (r))))
#define MUR_FMIX(_h) \
do { \
_h ^= _h >> 16; \
_h *= 0x85ebca6b; \
_h ^= _h >> 13; \
_h *= 0xc2b2ae35l; \
_h ^= _h >> 16; \
} while(0)
#define HASH_MUR(key,keylen,num_bkts,hashv,bkt) \
do { \
const uint8_t *_mur_data = (const uint8_t*)(key); \
const int _mur_nblocks = (keylen) / 4; \
uint32_t _mur_h1 = 0xf88D5353; \
uint32_t _mur_c1 = 0xcc9e2d51; \
uint32_t _mur_c2 = 0x1b873593; \
const uint32_t *_mur_blocks = (const uint32_t*)(_mur_data+_mur_nblocks*4); \
int _mur_i; \
for(_mur_i = -_mur_nblocks; _mur_i; _mur_i++) { \
uint32_t _mur_k1 = MUR_GETBLOCK(_mur_blocks,_mur_i); \
_mur_k1 *= _mur_c1; \
_mur_k1 = MUR_ROTL32(_mur_k1,15); \
_mur_k1 *= _mur_c2; \
\
_mur_h1 ^= _mur_k1; \
_mur_h1 = MUR_ROTL32(_mur_h1,13); \
_mur_h1 = _mur_h1*5+0xe6546b64; \
} \
const uint8_t *_mur_tail = (const uint8_t*)(_mur_data + _mur_nblocks*4); \
uint32_t _mur_k1=0; \
switch((keylen) & 3) { \
case 3: _mur_k1 ^= _mur_tail[2] << 16; \
case 2: _mur_k1 ^= _mur_tail[1] << 8; \
case 1: _mur_k1 ^= _mur_tail[0]; \
_mur_k1 *= _mur_c1; \
_mur_k1 = MUR_ROTL32(_mur_k1,15); \
_mur_k1 *= _mur_c2; \
_mur_h1 ^= _mur_k1; \
} \
_mur_h1 ^= (keylen); \
MUR_FMIX(_mur_h1); \
hashv = _mur_h1; \
bkt = hashv & (num_bkts-1); \
} while(0)
#endif /* HASH_USING_NO_STRICT_ALIASING */
/* key comparison function; return 0 if keys equal */
#define HASH_KEYCMP(a,b,len) memcmp(a,b,len)
/* iterate over items in a known bucket to find desired item */
#define HASH_FIND_IN_BKT(tbl,hh,head,keyptr,keylen_in,out) \
do { \
if (head.hh_head) DECLTYPE_ASSIGN(out,ELMT_FROM_HH(tbl,head.hh_head)); \
else out=NULL; \
while (out) { \
if ((out)->hh.keylen == keylen_in) { \
if ((HASH_KEYCMP((out)->hh.key,keyptr,keylen_in)) == 0) break; \
} \
if ((out)->hh.hh_next) DECLTYPE_ASSIGN(out,ELMT_FROM_HH(tbl,(out)->hh.hh_next)); \
else out = NULL; \
} \
} while(0)
/* add an item to a bucket */
#define HASH_ADD_TO_BKT(head,addhh) \
do { \
head.count++; \
(addhh)->hh_next = head.hh_head; \
(addhh)->hh_prev = NULL; \
if (head.hh_head) { (head).hh_head->hh_prev = (addhh); } \
(head).hh_head=addhh; \
if (head.count >= ((head.expand_mult+1) * HASH_BKT_CAPACITY_THRESH) \
&& (addhh)->tbl->noexpand != 1) { \
HASH_EXPAND_BUCKETS((addhh)->tbl); \
} \
} while(0)
/* remove an item from a given bucket */
#define HASH_DEL_IN_BKT(hh,head,hh_del) \
(head).count--; \
if ((head).hh_head == hh_del) { \
(head).hh_head = hh_del->hh_next; \
} \
if (hh_del->hh_prev) { \
hh_del->hh_prev->hh_next = hh_del->hh_next; \
} \
if (hh_del->hh_next) { \
hh_del->hh_next->hh_prev = hh_del->hh_prev; \
}
/* Bucket expansion has the effect of doubling the number of buckets
* and redistributing the items into the new buckets. Ideally the
* items will distribute more or less evenly into the new buckets
* (the extent to which this is true is a measure of the quality of
* the hash function as it applies to the key domain).
*
* With the items distributed into more buckets, the chain length
* (item count) in each bucket is reduced. Thus by expanding buckets
* the hash keeps a bound on the chain length. This bounded chain
* length is the essence of how a hash provides constant time lookup.
*
* The calculation of tbl->ideal_chain_maxlen below deserves some
* explanation. First, keep in mind that we're calculating the ideal
* maximum chain length based on the *new* (doubled) bucket count.
* In fractions this is just n/b (n=number of items,b=new num buckets).
* Since the ideal chain length is an integer, we want to calculate
* ceil(n/b). We don't depend on floating point arithmetic in this
* hash, so to calculate ceil(n/b) with integers we could write
*
* ceil(n/b) = (n/b) + ((n%b)?1:0)
*
* and in fact a previous version of this hash did just that.
* But now we have improved things a bit by recognizing that b is
* always a power of two. We keep its base 2 log handy (call it lb),
* so now we can write this with a bit shift and logical AND:
*
* ceil(n/b) = (n>>lb) + ( (n & (b-1)) ? 1:0)
*
*/
#define HASH_EXPAND_BUCKETS(tbl) \
do { \
unsigned _he_bkt; \
unsigned _he_bkt_i; \
struct UT_hash_handle *_he_thh, *_he_hh_nxt; \
UT_hash_bucket *_he_new_buckets, *_he_newbkt; \
_he_new_buckets = (UT_hash_bucket*)uthash_malloc( \
2 * tbl->num_buckets * sizeof(struct UT_hash_bucket)); \
if (!_he_new_buckets) { uthash_fatal( "out of memory"); } \
memset(_he_new_buckets, 0, \
2 * tbl->num_buckets * sizeof(struct UT_hash_bucket)); \
tbl->ideal_chain_maxlen = \
(tbl->num_items >> (tbl->log2_num_buckets+1)) + \
((tbl->num_items & ((tbl->num_buckets*2)-1)) ? 1 : 0); \
tbl->nonideal_items = 0; \
for(_he_bkt_i = 0; _he_bkt_i < tbl->num_buckets; _he_bkt_i++) \
{ \
_he_thh = tbl->buckets[ _he_bkt_i ].hh_head; \
while (_he_thh) { \
_he_hh_nxt = _he_thh->hh_next; \
HASH_TO_BKT( _he_thh->hashv, tbl->num_buckets*2, _he_bkt); \
_he_newbkt = &(_he_new_buckets[ _he_bkt ]); \
if (++(_he_newbkt->count) > tbl->ideal_chain_maxlen) { \
tbl->nonideal_items++; \
_he_newbkt->expand_mult = _he_newbkt->count / \
tbl->ideal_chain_maxlen; \
} \
_he_thh->hh_prev = NULL; \
_he_thh->hh_next = _he_newbkt->hh_head; \
if (_he_newbkt->hh_head) _he_newbkt->hh_head->hh_prev = \
_he_thh; \
_he_newbkt->hh_head = _he_thh; \
_he_thh = _he_hh_nxt; \
} \
} \
uthash_free( tbl->buckets, tbl->num_buckets*sizeof(struct UT_hash_bucket) ); \
tbl->num_buckets *= 2; \
tbl->log2_num_buckets++; \
tbl->buckets = _he_new_buckets; \
tbl->ineff_expands = (tbl->nonideal_items > (tbl->num_items >> 1)) ? \
(tbl->ineff_expands+1) : 0; \
if (tbl->ineff_expands > 1) { \
tbl->noexpand=1; \
uthash_noexpand_fyi(tbl); \
} \
uthash_expand_fyi(tbl); \
} while(0)
/* This is an adaptation of Simon Tatham's O(n log(n)) mergesort */
/* Note that HASH_SORT assumes the hash handle name to be hh.
* HASH_SRT was added to allow the hash handle name to be passed in. */
#define HASH_SORT(head,cmpfcn) HASH_SRT(hh,head,cmpfcn)
#define HASH_SRT(hh,head,cmpfcn) \
do { \
unsigned _hs_i; \
unsigned _hs_looping,_hs_nmerges,_hs_insize,_hs_psize,_hs_qsize; \
struct UT_hash_handle *_hs_p, *_hs_q, *_hs_e, *_hs_list, *_hs_tail; \
if (head) { \
_hs_insize = 1; \
_hs_looping = 1; \
_hs_list = &((head)->hh); \
while (_hs_looping) { \
_hs_p = _hs_list; \
_hs_list = NULL; \
_hs_tail = NULL; \
_hs_nmerges = 0; \
while (_hs_p) { \
_hs_nmerges++; \
_hs_q = _hs_p; \
_hs_psize = 0; \
for ( _hs_i = 0; _hs_i < _hs_insize; _hs_i++ ) { \
_hs_psize++; \
_hs_q = (UT_hash_handle*)((_hs_q->next) ? \
((void*)((char*)(_hs_q->next) + \
(head)->hh.tbl->hho)) : NULL); \
if (! (_hs_q) ) break; \
} \
_hs_qsize = _hs_insize; \
while ((_hs_psize > 0) || ((_hs_qsize > 0) && _hs_q )) { \
if (_hs_psize == 0) { \
_hs_e = _hs_q; \
_hs_q = (UT_hash_handle*)((_hs_q->next) ? \
((void*)((char*)(_hs_q->next) + \
(head)->hh.tbl->hho)) : NULL); \
_hs_qsize--; \
} else if ( (_hs_qsize == 0) || !(_hs_q) ) { \
_hs_e = _hs_p; \
_hs_p = (UT_hash_handle*)((_hs_p->next) ? \
((void*)((char*)(_hs_p->next) + \
(head)->hh.tbl->hho)) : NULL); \
_hs_psize--; \
} else if (( \
cmpfcn(DECLTYPE(head)(ELMT_FROM_HH((head)->hh.tbl,_hs_p)), \
DECLTYPE(head)(ELMT_FROM_HH((head)->hh.tbl,_hs_q))) \
) <= 0) { \
_hs_e = _hs_p; \
_hs_p = (UT_hash_handle*)((_hs_p->next) ? \
((void*)((char*)(_hs_p->next) + \
(head)->hh.tbl->hho)) : NULL); \
_hs_psize--; \
} else { \
_hs_e = _hs_q; \
_hs_q = (UT_hash_handle*)((_hs_q->next) ? \
((void*)((char*)(_hs_q->next) + \
(head)->hh.tbl->hho)) : NULL); \
_hs_qsize--; \
} \
if ( _hs_tail ) { \
_hs_tail->next = ((_hs_e) ? \
ELMT_FROM_HH((head)->hh.tbl,_hs_e) : NULL); \
} else { \
_hs_list = _hs_e; \
} \
_hs_e->prev = ((_hs_tail) ? \
ELMT_FROM_HH((head)->hh.tbl,_hs_tail) : NULL); \
_hs_tail = _hs_e; \
} \
_hs_p = _hs_q; \
} \
_hs_tail->next = NULL; \
if ( _hs_nmerges <= 1 ) { \
_hs_looping=0; \
(head)->hh.tbl->tail = _hs_tail; \
DECLTYPE_ASSIGN(head,ELMT_FROM_HH((head)->hh.tbl, _hs_list)); \
} \
_hs_insize *= 2; \
} \
HASH_FSCK(hh,head); \
} \
} while (0)
/* This function selects items from one hash into another hash.
* The end result is that the selected items have dual presence
* in both hashes. There is no copy of the items made; rather
* they are added into the new hash through a secondary hash
* hash handle that must be present in the structure. */
#define HASH_SELECT(hh_dst, dst, hh_src, src, cond) \
do { \
unsigned _src_bkt, _dst_bkt; \
void *_last_elt=NULL, *_elt; \
UT_hash_handle *_src_hh, *_dst_hh, *_last_elt_hh=NULL; \
ptrdiff_t _dst_hho = ((char*)(&(dst)->hh_dst) - (char*)(dst)); \
if (src) { \
for(_src_bkt=0; _src_bkt < (src)->hh_src.tbl->num_buckets; _src_bkt++) { \
for(_src_hh = (src)->hh_src.tbl->buckets[_src_bkt].hh_head; \
_src_hh; \
_src_hh = _src_hh->hh_next) { \
_elt = ELMT_FROM_HH((src)->hh_src.tbl, _src_hh); \
if (cond(_elt)) { \
_dst_hh = (UT_hash_handle*)(((char*)_elt) + _dst_hho); \
_dst_hh->key = _src_hh->key; \
_dst_hh->keylen = _src_hh->keylen; \
_dst_hh->hashv = _src_hh->hashv; \
_dst_hh->prev = _last_elt; \
_dst_hh->next = NULL; \
if (_last_elt_hh) { _last_elt_hh->next = _elt; } \
if (!dst) { \
DECLTYPE_ASSIGN(dst,_elt); \
HASH_MAKE_TABLE(hh_dst,dst); \
} else { \
_dst_hh->tbl = (dst)->hh_dst.tbl; \
} \
HASH_TO_BKT(_dst_hh->hashv, _dst_hh->tbl->num_buckets, _dst_bkt); \
HASH_ADD_TO_BKT(_dst_hh->tbl->buckets[_dst_bkt],_dst_hh); \
(dst)->hh_dst.tbl->num_items++; \
_last_elt = _elt; \
_last_elt_hh = _dst_hh; \
} \
} \
} \
} \
HASH_FSCK(hh_dst,dst); \
} while (0)
#define HASH_CLEAR(hh,head) \
do { \
if (head) { \
uthash_free((head)->hh.tbl->buckets, \
(head)->hh.tbl->num_buckets*sizeof(struct UT_hash_bucket)); \
HASH_BLOOM_FREE((head)->hh.tbl); \
uthash_free((head)->hh.tbl, sizeof(UT_hash_table)); \
(head)=NULL; \
} \
} while(0)
#ifdef NO_DECLTYPE
#define HASH_ITER(hh,head,el,tmp) \
for((el)=(head), (*(char**)(&(tmp)))=(char*)((head)?(head)->hh.next:NULL); \
el; (el)=(tmp),(*(char**)(&(tmp)))=(char*)((tmp)?(tmp)->hh.next:NULL))
#else
#define HASH_ITER(hh,head,el,tmp) \
for((el)=(head),(tmp)=DECLTYPE(el)((head)?(head)->hh.next:NULL); \
el; (el)=(tmp),(tmp)=DECLTYPE(el)((tmp)?(tmp)->hh.next:NULL))
#endif
/* obtain a count of items in the hash */
#define HASH_COUNT(head) HASH_CNT(hh,head)
#define HASH_CNT(hh,head) ((head)?((head)->hh.tbl->num_items):0)
typedef struct UT_hash_bucket {
struct UT_hash_handle *hh_head;
unsigned count;
/* expand_mult is normally set to 0. In this situation, the max chain length
* threshold is enforced at its default value, HASH_BKT_CAPACITY_THRESH. (If
* the bucket's chain exceeds this length, bucket expansion is triggered).
* However, setting expand_mult to a non-zero value delays bucket expansion
* (that would be triggered by additions to this particular bucket)
* until its chain length reaches a *multiple* of HASH_BKT_CAPACITY_THRESH.
* (The multiplier is simply expand_mult+1). The whole idea of this
* multiplier is to reduce bucket expansions, since they are expensive, in
* situations where we know that a particular bucket tends to be overused.
* It is better to let its chain length grow to a longer yet-still-bounded
* value, than to do an O(n) bucket expansion too often.
*/
unsigned expand_mult;
} UT_hash_bucket;
/* random signature used only to find hash tables in external analysis */
#define HASH_SIGNATURE 0xa0111fe1
#define HASH_BLOOM_SIGNATURE 0xb12220f2
typedef struct UT_hash_table {
UT_hash_bucket *buckets;
unsigned num_buckets, log2_num_buckets;
unsigned num_items;
struct UT_hash_handle *tail; /* tail hh in app order, for fast append */
ptrdiff_t hho; /* hash handle offset (byte pos of hash handle in element */
/* in an ideal situation (all buckets used equally), no bucket would have
* more than ceil(#items/#buckets) items. that's the ideal chain length. */
unsigned ideal_chain_maxlen;
/* nonideal_items is the number of items in the hash whose chain position
* exceeds the ideal chain maxlen. these items pay the penalty for an uneven
* hash distribution; reaching them in a chain traversal takes >ideal steps */
unsigned nonideal_items;
/* ineffective expands occur when a bucket doubling was performed, but
* afterward, more than half the items in the hash had nonideal chain
* positions. If this happens on two consecutive expansions we inhibit any
* further expansion, as it's not helping; this happens when the hash
* function isn't a good fit for the key domain. When expansion is inhibited
* the hash will still work, albeit no longer in constant time. */
unsigned ineff_expands, noexpand;
uint32_t signature; /* used only to find hash tables in external analysis */
#ifdef HASH_BLOOM
uint32_t bloom_sig; /* used only to test bloom exists in external analysis */
uint8_t *bloom_bv;
char bloom_nbits;
#endif
} UT_hash_table;
typedef struct UT_hash_handle {
struct UT_hash_table *tbl;
void *prev; /* prev element in app order */
void *next; /* next element in app order */
struct UT_hash_handle *hh_prev; /* previous hh in bucket order */
struct UT_hash_handle *hh_next; /* next hh in bucket order */
void *key; /* ptr to enclosing struct's key */
unsigned keylen; /* enclosing struct's key len */
unsigned hashv; /* result of hash-fcn(key) */
} UT_hash_handle;
#endif /* UTHASH_H */

3
include/emdb.hrl → include/lmdb.hrl
View file

@ -1,7 +1,8 @@
%%-------------------------------------------------------------------
%% This file is part of EMDB - Erlang MDB API
%% This file is part of LMDB - Erlang Lightning MDB API
%%
%% Copyright (c) 2012 by Aleph Archives. All rights reserved.
%% Copyright (c) 2013 by Basho Technologies, Inc. All rights reserved.
%%
%%-------------------------------------------------------------------
%% Redistribution and use in source and binary forms, with or without

BIN
rebar vendored
View file

Binary file not shown.

14
rebar.config
View file

@ -5,9 +5,8 @@
{cover_enabled, true}.
{erl_opts, [{d,'DEBUG',true},
{erl_opts, [%{d,'DEBUG',true},
debug_info,
fail_on_warning,
warn_unused_vars,
warn_export_all,
warn_shadow_vars,
@ -21,7 +20,8 @@
warn_exported_vars,
warn_untyped_record
%warn_missing_spec,
%strict_validation
%strict_validation,
%fail_on_warning
]}.
{xref_checks, [undefined_function_calls, deprecated_function_calls]}.
@ -29,10 +29,10 @@
{eunit_opts, [verbose, {report, {eunit_surefire, [{dir, "."}]}}]}.
{port_specs, [
{"unix", "priv/emdb.so", ["c_src/*.c"]},
{"linux", "priv/emdb.so", ["c_src/*.c"]},
{"darwin", "priv/emdb.so", ["c_src/*.c"]},
{"win32", "priv/emdb.dll", ["c_src/*.c"]}
{"unix", "priv/lmdb.so", ["c_src/*.c"]},
{"linux", "priv/lmdb.so", ["c_src/*.c"]},
{"darwin", "priv/lmdb.so", ["c_src/*.c"]},
{"win32", "priv/lmdb.dll", ["c_src/*.c"]}
]}.
{port_env, [

22
src/async_nif.hrl
View file

@ -1,25 +1,23 @@
%% -------------------------------------------------------------------
%% ---------------------------------------------------------------------------
%%
%% async_nif: An async thread-pool layer for Erlang's NIF API
%%
%% Copyright (c) 2012 Basho Technologies, Inc. All Rights Reserved.
%% Copyright (c) 2012-2013 Basho Technologies, Inc. All Rights Reserved.
%% Author: Gregory Burd <greg@basho.com> <greg@burd.me>
%%
%% This file is provided to you under the Apache License,
%% Version 2.0 (the "License"); you may not use this file
%% except in compliance with the License. You may obtain
%% a copy of the License at
%% This file is provided to you under the Apache License, Version 2.0 (the
%% "License"); you may not use this file except in compliance with the License.
%% You may obtain a copy of the License at:
%%
%% http://www.apache.org/licenses/LICENSE-2.0
%%
%% Unless required by applicable law or agreed to in writing,
%% software distributed under the License is distributed on an
%% "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
%% KIND, either express or implied. See the License for the
%% specific language governing permissions and limitations
%% Unless required by applicable law or agreed to in writing, software
%% distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
%% WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
%% License for the specific language governing permissions and limitations
%% under the License.
%%
%% -------------------------------------------------------------------
%% ---------------------------------------------------------------------------
-spec async_nif_enqueue(reference(), function(), [term()]) -> term() | {error, term()}.
async_nif_enqueue(R, F, A) ->

6
src/emdb.app.src → src/lmdb.app.src
View file

@ -1,7 +1,7 @@
{application, emdb,
{application, lmdb,
[
{description, "Memory-Mapped Database"},
{vsn, "0.9.0"},
{description, "OpenLDAP's Lightning Memory-Mapped Database BTREE"},
{vsn, "1.0.0"},
{registered, []},
{applications, [
kernel,

103
src/emdb.erl → src/lmdb.erl
View file

@ -1,7 +1,8 @@
%%-------------------------------------------------------------------
%% This file is part of EMDB - Erlang MDB API
%% This file is part of LMDB - Erlang Lightning MDB API
%%
%% Copyright (c) 2012 by Aleph Archives. All rights reserved.
%% Copyright (c) 2013 by Basho Technologies, Inc. All rights reserved.
%%
%%-------------------------------------------------------------------
%% Redistribution and use in source and binary forms, with or without
@ -25,13 +26,13 @@
%% OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
%%-------------------------------------------------------------------
-module(emdb).
-module(lmdb).
%%====================================================================
%% EXPORTS
%%====================================================================
-export([
open/1,
%open/1,
open/2,
open/3,
@ -52,7 +53,7 @@
%% config for testing
-ifdef(TEST).
-ifdef(EQC).
%include_lib("eqc/include/eqc.hrl").
-include_lib("eqc/include/eqc.hrl").
-define(QC_OUT(P), eqc:on_output(fun(Str, Args) -> io:format(user, Str, Args) end, P)).
-endif.
-include_lib("eunit/include/eunit.hrl").
@ -61,15 +62,15 @@
%%====================================================================
%% Includes
%%====================================================================
-include("emdb.hrl").
-include("lmdb.hrl").
-include("async_nif.hrl").
%%====================================================================
%% MACROS
%%====================================================================
-define(EMDB_DRIVER_NAME, "emdb").
-define(LMDB_DRIVER_NAME, "lmdb").
-define(NOT_LOADED, not_loaded(?LINE)).
-define(MDB_MAP_SIZE, 10485760). %% 10MB
-define(MDB_MAP_SIZE, 2147483648). %% 2GB in bytes
%%====================================================================
%% PUBLIC API
@ -79,8 +80,8 @@
%% @doc Create a new MDB database
%% @end
%%--------------------------------------------------------------------
open(DirName) ->
open(DirName, ?MDB_MAP_SIZE).
%open(DirName) ->
% open(DirName, ?MDB_MAP_SIZE).
open(DirName, MapSize)
when is_integer(MapSize)
andalso MapSize > 0 ->
@ -155,7 +156,7 @@ init() ->
Path ->
Path
end,
erlang:load_nif(filename:join(PrivDir, ?EMDB_DRIVER_NAME), 0).
erlang:load_nif(filename:join(PrivDir, ?LMDB_DRIVER_NAME), 0).
%%--------------------------------------------------------------------
@ -170,20 +171,21 @@ not_loaded(Line) ->
%% ===================================================================
-ifdef(TEST).
-define(TEST_DATA_DIR, "test/basics").
open_test_db(DataDir) ->
open_test_db() ->
{ok, CWD} = file:get_cwd(),
Path = filename:join([CWD, DataDir]),
?cmd("rm -rf " ++ Path),
?assertMatch(ok, filelib:ensure_dir(filename:join([Path, "x"]))),
{ok, Handle} = ?MODULE:open(Path),
DataDir = filename:join([CWD, "test", "eunit"]),
?cmd("rm -rf " ++ DataDir),
?assertMatch(ok, filelib:ensure_dir(filename:join([DataDir, "x"]))),
{ok, Handle} = ?MODULE:open(DataDir, 2147483648),
[?MODULE:upd(Handle, crypto:sha(<<X>>),
crypto:rand_bytes(crypto:rand_uniform(128, 4096))) ||
X <- lists:seq(1, 100)],
Handle.
basics_test_() ->
{setup,
fun() ->
open_test_db(?TEST_DATA_DIR)
open_test_db()
end,
fun(Handle) ->
ok = ?MODULE:close(Handle)
@ -192,16 +194,16 @@ basics_test_() ->
{inorder,
[{"open and close a database",
fun() ->
Handle = open_test_db(Handle)
Handle = open_test_db()
end},
{"create, then drop an empty database",
fun() ->
Handle = open_test_db(Handle),
Handle = open_test_db(),
?assertMatch(ok, ?MODULE:drop(Handle))
end},
{"create, put an item, get it, then drop the database",
fun() ->
Handle = open_test_db(Handle),
Handle = open_test_db(),
?assertMatch(ok, ?MODULE:put(Handle, <<"a">>, <<"apple">>)),
?assertMatch(ok, ?MODULE:put(Handle, <<"b">>, <<"boy">>)),
?assertMatch(ok, ?MODULE:put(Handle, <<"c">>, <<"cat">>)),
@ -215,4 +217,63 @@ basics_test_() ->
]}
end}.
-ifdef(EQC).
qc(P) ->
?assert(eqc:quickcheck(?QC_OUT(P))).
keys() ->
eqc_gen:non_empty(list(eqc_gen:non_empty(binary()))).
values() ->
eqc_gen:non_empty(list(binary())).
ops(Keys, Values) ->
{oneof([put, delete]), oneof(Keys), oneof(Values)}.
apply_kv_ops([], _Handle, Acc0) ->
Acc0;
apply_kv_ops([{put, K, V} | Rest], Handle, Acc0) ->
ok = ?MODULE:put(Handle, K, V),
apply_kv_ops(Rest, Handle, orddict:store(K, V, Acc0));
apply_kv_ops([{del, K, _} | Rest], Handle, Acc0) ->
ok = case ?MODULE:del(Handle, K) of
ok ->
ok;
not_found ->
ok;
Else ->
Else
end,
apply_kv_ops(Rest, Handle, orddict:store(K, deleted, Acc0)).
prop_put_delete() ->
?LET({Keys, Values}, {keys(), values()},
?FORALL(Ops, eqc_gen:non_empty(list(ops(Keys, Values))),
begin
{ok, CWD} = file:get_cwd(),
DataDir = filename:join([CWD, "test", "eqc"]),
?cmd("rm -rf " ++ DataDir),
ok = filelib:ensure_dir(filename:join([DataDir, "x"])),
{ok, Handle} = ?MODULE:open(DataDir, 2147483648),
try
Model = apply_kv_ops(Ops, Handle, []),
%% Validate that all deleted values return not_found
F = fun({K, deleted}) ->
?assertEqual(not_found, ?MODULE:get(Handle, K));
({K, V}) ->
?assertEqual({ok, V}, ?MODULE:get(Handle, K))
end,
lists:map(F, Model),
true
after
?MODULE:close(Handle)
end
end)).
prop_put_delete_test_() ->
{timeout, 3*60, fun() -> qc(prop_put_delete()) end}.
-endif.
-endif.

8
src/emdb_sup.erl → src/lmdb_sup.erl
View file

@ -1,7 +1,7 @@
%%-------------------------------------------------------------------
%% This file is part of EMDB - Erlang MDB API
%% This file is part of LMDB - Erlang Lightning MDB API
%%
%% Copyright (c) 2013 by Basho Technologies. All rights reserved.
%% Copyright (c) 2013 by Basho Technologies, Inc. All rights reserved.
%%
%%-------------------------------------------------------------------
%% Redistribution and use in source and binary forms, with or without
@ -25,7 +25,7 @@
%% OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
%%-------------------------------------------------------------------
-module(emdb_sup).
-module(lmdb_sup).
-behaviour(supervisor).
@ -53,4 +53,4 @@ start_link() ->
%% Supervisor callbacks
%% ===================================================================
init([]) ->
{ok, {{one_for_one, 5, 10}, [?CHILD(emdb, worker)]}}.
{ok, {{one_for_one, 5, 10}, [?CHILD(lmdb, worker)]}}.

24
tools/basho_bench_driver_emdb.erl → tools/basho_bench_driver_lmdb.erl
View file

@ -1,4 +1,4 @@
-module(basho_bench_driver_emdb).
-module(basho_bench_driver_lmdb).
-record(state, {
handle
@ -15,35 +15,35 @@
%% ====================================================================
new(1) ->
%% Make sure emdb is available
case code:which(emdb) of
%% Make sure lmdb is available
case code:which(lmdb) of
non_existing ->
?FAIL_MSG("~s requires emdb to be available on code path.\n",
?FAIL_MSG("~s requires lmdb to be available on code path.\n",
[?MODULE]);
_ ->
ok
end,
%{ok, _} = emdb_sup:start_link(),
%{ok, _} = lmdb_sup:start_link(),
setup(1);
new(Id) ->
setup(Id).
setup(_Id) ->
%% Get the target directory
Dir = basho_bench_config:get(emdb_dir, "/tmp"),
%Config = basho_bench_config:get(emdb, []),
Dir = basho_bench_config:get(lmdb_dir, "/tmp"),
%Config = basho_bench_config:get(lmdb, []),
%% Start Lightning MDB
case emdb:open(Dir, 32212254720, 16#10000 bor 16#40000 bor 16#80000) of
case lmdb:open(Dir, 32212254720, 16#10000 bor 16#40000 bor 16#80000) of
{ok, H} ->
{ok, #state{handle=H}};
{error, Reason} ->
?FAIL_MSG("Failed to establish a Lightning MDB connection, emdb backend unable to start: ~p\n", [Reason]),
?FAIL_MSG("Failed to establish a Lightning MDB connection, lmdb backend unable to start: ~p\n", [Reason]),
{error, Reason}
end.
run(get, KeyGen, _ValueGen, #state{handle=Handle}=State) ->
case emdb:get(Handle, KeyGen()) of
case lmdb:get(Handle, KeyGen()) of
{ok, _Value} ->
{ok, State};
not_found ->
@ -54,14 +54,14 @@ run(get, KeyGen, _ValueGen, #state{handle=Handle}=State) ->
run(put, KeyGen, ValueGen, #state{handle=Handle}=State) ->
Key = KeyGen(),
Val = ValueGen(),
case emdb:upd(Handle, Key, Val) of
case lmdb:upd(Handle, Key, Val) of
ok ->
{ok, State};
{error, Reason} ->
{error, Reason}
end;
run(delete, KeyGen, _ValueGen, #state{handle=Handle}=State) ->
case emdb:del(Handle, KeyGen()) of
case lmdb:del(Handle, KeyGen()) of
ok ->
{ok, State};
not_found ->

26
tools/emdb.config → tools/lmdb.config
View file

@ -2,23 +2,23 @@
%% ex: ft=erlang ts=4 sw=4 et
%% How to:
%% * put the emdb.config file into basho_bench/examples
%% * put the basho_bench_driver_emdb.erl into basho_bench/src
%% * put the lmdb.config file into basho_bench/examples
%% * put the basho_bench_driver_lmdb.erl into basho_bench/src
%% * make clean in basho_bench, then make
%% * edit examples/emdb.config
%% - change {code_paths, ["../emdb"]}. to be a relative path to your
%% emdb directory
%% - change {emdb_dir, "/home/gburd/ws/basho_bench/data"}. to a fully
%% * edit examples/lmdb.config
%% - change {code_paths, ["../lmdb"]}. to be a relative path to your
%% lmdb directory
%% - change {lmdb_dir, "/home/gburd/ws/basho_bench/data"}. to a fully
%% qualified location for your test data files (mkdir that directory
%% yourself, if it doesn't exist the test will fail 'enoent')
%% * to run, replace this path with the proper path on your system:
%% LD_LIBRARY_PATH=/home/you/emdb/priv ./basho_bench examples/emdb.config
%% LD_LIBRARY_PATH=/home/you/lmdb/priv ./basho_bench examples/lmdb.config
%% * the test should run for 10 minutes (as it is configured right now)
%% with 4 concurrent workers accessing the same table
%%
%% Note:
%% There are two config sections in wt.config {emdb, [ ... ]}. and
%% {emdb_, [ ... ]}. The one being used is named "emdb" the other
%% There are two config sections in wt.config {lmdb, [ ... ]}. and
%% {lmdb_, [ ... ]}. The one being used is named "lmdb" the other
%% config is ignored. I setup an LSM and BTREE config and to choose
%% which is run you just rename those two sections (turn one off by
%% adding a "_" to the name and take the "_" out of the other's name).
@ -26,12 +26,12 @@
{mode, max}.
{duration, 10}.
{concurrent, 8}.
{driver, basho_bench_driver_emdb}.
{driver, basho_bench_driver_lmdb}.
{key_generator, {int_to_bin_littleendian,{uniform_int, 5000000}}}.
{value_generator, {fixed_bin, 1024}}.
%{operations, [{get, 9}, {put, 9}, {delete, 2}]}.
{operations, [{put, 1}]}.
{code_paths, ["../emdb"]}.
{emdb_dir, "/home/gburd/ws/basho_bench/data"}.
{code_paths, ["../lmdb"]}.
{lmdb_dir, "/home/gburd/ws/basho_bench/data"}.
{emdb, [ ]}.
{lmdb, [ ]}.