The hb-ddb from teamhackback

hb-ddb's Issues

Make properly @safe

Fix the holes in @trusted.

Support copy in/out for fast binary data storage

Always send binary data for bind messages

                        case BOOLEAN:
                        case TIMESTAMP:
                        case INET:
                        case NUMERIC:
                        case JSONB:
                        case INTERVAL:
                        case VARCHAR:
                        case TEXT:
                            stream.write(cast(short) 0); // text format
                            break;
                        default:
                            stream.write(cast(short) 1); // binary format
                    }

PGCommand:executeQuery -> rename to query

Should be the same as in PGConnection, but here query is a variable an function already

Evaluate when to use the extended protocol (binary)

requires two trips
simple one is always text
maybe simple one is faster for small queries or commands (listen..)
other projects switched to extended protocol only:

npgsql/npgsql#447

Implement HStore

Avoid the need to query for all OIds in the beginning

separate serialization logic

It's currently split into multiple bits

Fix package(db) visibilities

-> proper encapsulation

Handle ReadyForQuery directly

and maybe even Async ones?

Performance check (memory allocation)

use e.g. FreeListRef for scoped allocation of classes

Postgres interval mapping

interval is converted to Duration, which does not support months

PGConnection: executeQuery without PGCommand

Currently executeQuery (or executeRow, executeNonQuery or executeScalar) create a temporary PGCommand`:

PGResultSet!Specs executeQuery(Specs...)(string query)
{
    scope cmd = FreeListRef!PGCommand(this, query);
    return cmd.executeQuery!Specs();
}

which itself calls:

PGResultSet!Specs executeQuery(Specs...)()
{
    checkPrepared(true);
    checkBound();
    return conn.executeQuery!Specs(preparedName, _fields);
}

Maybe it might make sense to avoid creating a prepared statement at all?

Add performance benchmark

Json data

Add executeScalar (from PGCommand) to PGConnection

CTFE optimizations

genLabels adds a huge overhead to the compile-time.

Handle big messages

Messages.type: use MessageType

Data types: Arrays

PGMessages: extend support for plain text mode

Plain text mode is used by simple queries, but not very well supported.

PGConnection: remove result.destroy

Deprecated as ResultSet is now ref-counted

Handle parse errors

         void sendParseMessage(string statementName, string query, int[] oids)
         {
             int len = cast(int)(4 + statementName.length + 1 + query.length + 1 + 2 + oids.length * 4);		           
-		+			bool failed;
-            stream.write('P');		+            try stream.write('P');
+			catch (Exception e) {
+				failed = true;
+			}
+			if (failed) {
+				close();
+				throw new Exception("Error in parse: " ~ statementName ~ " query: " ~ query);
+			}

Use PostgresConnectionSettings struct instead of dict

Like http://vibed.org/api/vibe.db.mongo.settings/MongoClientSettings

Add debug mode: show SQL messages

Exceptions: proper inheritance + file, line overloads

Inherit from PGException

PreparedStatements: parameter handling

PGStream: support unix socket connection

this(UDSConnection socket)

PGConnection: add `createCommand`

Shortcut to create new commands. May use FreeListRef

Check ConnectionPool

Does it make sense to release an already connected & authorized stream?
A: No, it gets back to the connection pool
Test connection reuse

Add test suite

SSL support

Something similar like:

auto conn = connectTCP(m_options.address, m_options.port);
auto sslctx = new SSLContext(m_options.cert, m_options.key);
auto stream = new SSLStream(conn, sslctx, SSLStreamState.connecting);