In an attempt to learn D and git I'm building a Redis client. Assertedly
there are many problems in the implementation, I'd like to hear opinions
pointing them or suggestions. If here isn't the place to ask these type of
review, where is?
github.com/pslacerda/dredis

Some friend asked why I choose not to go asynchronous. It's because usually clients and servers are at a very fast connection and Redis responses are pretty fast too.

Correct me if wrong, but I need to do nothing special at client.RedisClient to make it thread safe, right?

token.__Bulk has an nullable array because I made distinction between an empty and a null arrays as the protocol did. The type looks fully bad. Some idea?

I guess that make token.Token an struct and pass it through ref functions will make the code somewhat better, but I have no clues. Token will never be seen by the user. When should I do it?

I guess that implement the protocol as free functions isn't a bad option because D offers this imperative style and the protocol is so small that doesn't need more a structured approach. Tell me this is wrong!


thanks from a newbie :-)
Pedro Lacerda

On 05.03.2012 7:24, Pedro Lacerda wrote:
> In an attempt to learn D and git I'm building a Redis client. Assertedly
> there are many problems in the implementation, I'd like to hear opinions
> pointing them or suggestions. If here isn't the place to ask these type
> of review, where is?
> github.com/pslacerda/dredis <http://github.com/pslacerda/dredis>

You may go for D group, though for newbie kind of thing D.learn is perfect place.

>
> Some friend asked why I choose not to go asynchronous. It's because
> usually clients and servers are at a very fast connection and Redis
> responses are pretty fast too.

Right, and you always can go fancy later on :)

>
> Correct me if wrong, but I need to do nothing special at
> client.RedisClient to make it thread safe, right?
>
In a sense. I mean everything is thread-local by default so unless you plan to go _shared_ it's OK.

> token.__Bulk has an nullable array because I made distinction between an
> empty and a null arrays as the protocol did. The type looks fully bad.
> Some idea?

Just use plain Nullable!(T[]) as is? I mean there is no need to wrap wrapper type ;)

>
> I guess that make token.Token an struct and pass it through ref
> functions will make the code somewhat better, but I have no clues. Token
> will never be seen by the user. When should I do it?

Make it a struct, from the looks of it it's a tagged union anyway. More importantly passing it _by value_ and copying is far cheaper then allocating new objects and working with them via extra indirection.
It would be almost the same as using Algebraic!(string,Exception,long,Bulk,MultiBulk) from std.variant. Though then you'd miss your nice Type enumeration.

>
> I guess that implement the protocol as free functions isn't a bad option
> because D offers this imperative style and the protocol is so small that
> doesn't need more a structured approach. Tell me this is wrong!
>

It's up to you in any case. Benefits of extra structure won't show up until the code is quite large.

>
> thanks from a newbie :-)
> Pedro Lacerda

You are welcome !

-- 
Dmitry Olshansky