I forked your repo to DocSavage/gomdb and put what I think are the necessary mdb files in a src/ directory. Since the code base is fairly small and pure C, I thought it would allow static linking and a simple exe.

Unfortunately, I'm getting the following:

# github.com/DocSavage/gomdb
In file included from ./env.go:10:
./src/mdb.c:8404:46: warning: data argument not used by format string [-Wformat-extra-args]
/usr/include/secure/_stdio.h:47:56: note: expanded from macro 'sprintf'
# github.com/DocSavage/gomdb
./txn.go:58: cannot use env._env (type *_Ctype_MDB_env) as type *[0]byte in function argument

The issue with txn.go doesn't make sense but this is my first time doing cgo bindings. It looks like it's properly declared. Not sure where the *[0]byte signature is coming from. Any suggestions?

I am trying to do a open, set, close and open get close based on github.com/cangus/phase example.go

What can I change to run this correctly correctly? i.e. reopen a prior environment and retrieve a key? Thanks.

https://gist.github.com/jcrubino/2b17210d986ab2e223ee

Hi,

gomd is a dependency of influxdb and I was unable to compile it on freebsd 10.

./gomdb/mdb.c:4300:39: error: use of undeclared identifier 'O_DSYNC'
../gomdb/mdb.c:282:20: note: expanded from macro 'MDB_DSYNC'
../gomdb/mdb.c:8584:46: warning: data argument not used by format string [-Wformat-extra-args]

O_DSYNC is not defined under FreeBSD according to my friend Google. I am not familar with C/C++ code, hence I didn't submit a pull request, but the following snippet seems to fix it for me.

#ifndef MDB_DSYNC
#ifdef O_DSYNC
# define MDB_DSYNC  O_DSYNC
#else
# define MDB_DSYNC      O_SYNC
#endif

Again, I have no idea what I'm doing - but this worked for me. Was that the right thing to do?

After building LMDB from openldap/libraries/liblmdb and settingd LD path LD_LIBRARY_PATH=/usr/local/lib I get a compilation error on Mac OSX 10.9 (Mavericks).

A simple go get github.com/szferi/gomdb will yield

# github.com/szferi/gomdb
clang: error: argument unused during compilation: '-fno-eliminate-unused-debug-types'

GCC Version

gcc --version
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.9.sdk/usr/include/c++/4.2.1
Apple LLVM version 5.0 (clang-500.2.79) (based on LLVM 3.3svn)
Target: x86_64-apple-darwin13.0.0
Thread model: posix

gcc-42 --version                                                                                                         
i686-apple-darwin11-gcc-4.2.1 (GCC) 4.2.1 (Apple Inc. build 5666) (dot 3)

Commit #206e1af broke the tests:

# github.com/szferi/gomdb
./bench_test.go:159: not enough arguments in call to cur.Get
./bench_test.go:210: not enough arguments in call to cur.GetVal
./example_test.go:46: not enough arguments in call to cursor.Get
./mdb_test.go:85: not enough arguments in call to cursor.Get
FAIL    github.com/szferi/gomdb [build failed]

Are there any plans to upgrade the lmdb implementation to 0.9.17 or is this project essentially frozen?

Hi, I'd be very interested in using this implementation of lmdb, and I was wondering if you could paste a usage example in the README at least.

[root@localhost gomdb]# go version
go version go1.8 linux/amd64

[root@localhost gomdb]# go test -bench .
--- FAIL: TestTest1 (0.00s)
panic: runtime error: cgo argument has Go pointer to Go pointer [recovered]
	panic: runtime error: cgo argument has Go pointer to Go pointer

goroutine 23 [running, locked to thread]:
testing.tRunner.func1(0xc4200789c0)
	/usr/local/go/src/testing/testing.go:622 +0x29d
panic(0x543e00, 0xc420070f80)
	/usr/local/go/src/runtime/panic.go:489 +0x2cf
github.com/szferi/gomdb.(*Txn).Put.func1(0x217ce960, 0xc400000001, 0xc420070f60, 0xc420070f70, 0xc400000010, 0x5)
	/home/go-project/src/github.com/szferi/gomdb/txn.go:149 +0x8b
github.com/szferi/gomdb.(*Txn).Put(0xc4200820d8, 0x1, 0xc420070f50, 0x5, 0x8, 0xc420070f58, 0x5, 0x8, 0x10, 0x0, ...)
	/home/go-project/src/github.com/szferi/gomdb/txn.go:149 +0xf7
github.com/szferi/gomdb.TestTest1(0xc4200789c0)
	/home/go-project/src/github.com/szferi/gomdb/mdb_test.go:52 +0x7ec
testing.tRunner(0xc4200789c0, 0x572d90)
	/usr/local/go/src/testing/testing.go:657 +0x96
created by testing.(*T).Run
	/usr/local/go/src/testing/testing.go:697 +0x2ca
exit status 2
FAIL	github.com/szferi/gomdb	0.006s

The MDB library does not define errors for mdb_env_info, but Env.Info() returns an error. This is overly pedantic. Env.Info() should return just the Info type.

edit: I see that the function can return EINVAL when either pointer passed is nil. But that's be an acceptable place to panic in these binding, imo. The *Info pointer will never be nil. And calling methods on nil values is not a common practice in Go. Though it does work... I don't call things on nil values unless the documentation says it's ok.

these functions don't really seem to fit in the package. they are somewhat high-level for an otherwise low-level api. and, really, no one uses gob in practice (afaik).

would you mind if i opened a pull request to remove them?

root@aa685bdc1a0f:/gopath/src/github.com/szferi/gomdb# go build

github.com/szferi/gomdb

./mdb.c: In function 'mdb_env_open':
./mdb.c:4481:39: error: 'O_DSYNC' undeclared (first use in this function)
./mdb.c:4481:39: note: each undeclared identifier is reported only once for each
./mdb.c: In function 'mdb_env_copyfd1':
./mdb.c:8497:2: warning: implicit declaration of function 'posix_memalign' [-Wim
root@aa685bdc1a0f:/gopath/src/github.com/szferi/gomdb#

GOARCH =arm
GOOS=android

FYI, I'm getting these build warnings on Mac OS 10.9.4, go version 1.3.1:

github.com/szferi/gomdb/mdb.c:5862:9: warning: variable 'rc' is used uninitialized whenever 'if' condition is false [-Wsometimes-uninitialized]
github.com/szferi/gomdb/mdb.c:5888:9: note: uninitialized use occurs here
github.com/szferi/gomdb/mdb.c:5862:5: note: remove the 'if' if its condition is always true
github.com/szferi/gomdb/mdb.c:5721:10: note: initialize the variable 'rc' to silence this warning
github.com/szferi/gomdb/mdb.c:9260:46: warning: data argument not used by format string [-Wformat-extra-args]

The warnings don't seem serious, but is this something we should consider fixing in mdb.c and pushing upstream to MDB's maintainer?

The usage of runtime.LockOSThread() and runtime.UnlockOSThread() seems incorrect to me.

Child transactions will unlock their parents on Commit or Abort. Dealing with this is fairly trivial except maybe in the case of Cursor.Txn, where it seem to get only slightly more tricky.

The bigger problem in my opinion seems to be that there's nothing preventing usage of the transaction, or its cursors, inside another goroutine which, due to thread locking, is guaranteed by the Go runtime to execute on another OS thread.

I see two ways to deal with these problems.

• Fix the problem with child transactions and place warning in godoc stating that write transactions (those without mdb.RDONLY) absolutely must not be used in any goroutine other than the one that created it. (edit: pull request #31)
• Abstract this problem away by creating a new, locked goroutine in Env.BeginTxn and serialize all operations through it via channels. This would include serializing cursor creation and use through the same goroutine. Because MDB_NOTLS is used this would only be required for write transactions.

The latter option will introduce some overhead into write transactions. But I'm starting to think it's the only sane way to use MDB in Go.

Do you have any thoughts on this @szferi? I've spent a good amount of time looking into this but wouldn't be surprised if I've misinterpreted the rather spartan C docs. So please validate my conclusions yourself.

Hello, in my PoC trying to using the gomdb as system of persistence in memory, it's generated a error "MDB_NOTFOUND: No matching key/data pair found"

1. The data are loading in memory through gomdb
2. in the main.go is created the instance the "env *mdb.Env", other component(package) is invoked to load data through method with param "env"
3. the load data is successful.
4. when the service REST is invoked, try to search the information in memory through the instance "env (gomdb)" and generates the error.
5. In the service is created a TRX and DBI instance, open one Cursor and use the method Get " _, bval, err := cursor.Get([]byte("H0"), mdb.SET_KEY)) "

//
txn, err := env.BeginTxn(nil, 0)
if err != nil {
fmt.Printf("Cannot begin transaction: %s", err)
}
var dbi mdb.DBI
dbi, err = txn.DBIOpen(nil, 0)
defer env.DBIClose(dbi)

Can you tell me about is generated this error or one example to use different to the test_case in the project

[gnewton@chekov]$ go get github.com/szferi/gomdb

github.com/szferi/gomdb

txn.go:146[/tmp/go-build877661856/github.com/szferi/gomdb/_obj/txn.cgo1.go:166]: val redeclared in this block
previous declaration at txn.go:138[/tmp/go-build877661856/github.com/szferi/gomdb/_obj/txn.cgo1.go:158]
[gnewton@chekov]$

Fixed doing the following (changed 'val' to 'bval' from line that was previously: var val []byte:

func (txn *Txn) GetGo(dbi DBI, key, val interface{}) error {
var key_buffer bytes.Buffer
encoder := gob.NewEncoder(&key_buffer)
err := encoder.Encode(key)
if err != nil {
return err
}
gkey := key_buffer.Bytes()
var bval []byte
bval, err = txn.Get(dbi, gkey)
if err != nil {
return err
}
val_buffer := bytes.NewReader(bval)
decoder := gob.NewDecoder(val_buffer)
err = decoder.Decode(bval)
if err != nil {
return err
}
return nil
}

Builds and all tests pass with this change.