Author here. It's funny, I actually didn't even know about HoneySQL when I made this. It was only after that someone pointed it out to me that I saw how similar our libraries are.
It believe it says something really good about this approach that we both came to such a similar conclusion independently.
I think OJ takes a slightly more minimalistic approach. I would say that my specification is a little more precise as well. Maybe in the future the author of HoneySQL and I can collaborate.
+1 - This is by far the sanest approach to rdbms interaction I've seen. It's kind of sad that we're stuck with SQL but while we're stuck with it we should stick to it. :P
I agree and it's what I've used for all of my pet clojure projects that need to talk to a database. However, the drawback it suffers versus a lot of the approaches taking by most of the alternative libraries is that it doesn't provide any way to compose queries.
I'm talking about the kind of stuff something like Korma or SQLAlchemy enable where you build up a query to be executed dynamically based on, for example, user filters.
I haven't need anything like that on any of my pet projects so I've been deliriously happy with Yesql. But, while working on those pet projects I do often find myself thinking about whatever I'm doing in my day job and I haven't come up with a good way to support composability within the context of Yesql. I would love to see a solution to that.
1. Queries are executed by passing your query and db-config to oj/exec. So, does it mean that with any new query there will be new connection to db? How do you manage transaction, when doing inserts/updates? I suppose there must be some magic I don't understand, because if that stuff actually is unsupported that would make all thing pretty much unusable, wouldn't it? Well, if you are not using MySQL/MyISAM anyway.
2. I suppose that library does some internal sanitizing, because it is what such libraries usually are for after all. There isn't much sense in using keywords instead of strings if it is the same plain SQL you would write manually (well, you wouldn't make a typo in the word WHERE, but that's all). But I don't see any mentions of types whatsoever. So if I pass some variable foo to that query, how does it know how to validate and sanitize it? If it is allowed to be NULL? If it should be passed to query as INT or string or maybe DATETIME? As it is minimalistic I would assume it doesn't use schema (and how would it if it's not an ORM, but just some query composer?), so how does it work then?
There are few other things I frowned on, but these are the major points.
I'll try to address your concerns, but keep in mind: this is a very new project.
1. When you "exec" a query, all I'm really doing is compiling the query map into an SQL statement and passing it into java.jdbc[1] with the database configuration that is given. This database config can be stored as a regular Clojure map in a config namespace in your project or however you wish. To answer your question, yes, at the moment java.jdbc is called for each query that is executed and I can only assume that that means that a new connection is made. Obviously there is room for improvement here, which will be addressed in the future.
2. I've provided a thorough specification for what makes a Clojure map a "query map" in the SPEC file[2]. Before any query map is compiled into SQL, it gets validated by a function that ensures that it meets the specification. If not, it throws an error.
Hopefully this answers some of your concerns! Like I said, this is a new project that I made for fun and for personal use. I plan on working on it for a long time, and I'm sure that many things will be improved! I hope that others find it useful (or at least interesting) as well.
Regarding point 1, java.jdbc can accept a variety of types of connection maps. The simplest scenario is you pass in a map that contains the connection information, and you get a new connection and transaction for each call. However, you can also pass in live connections/transactions that can be reused. The with-db-transaction form is one way to accomplish this. It's really pretty flexible.
(And btw, your library looks really quite good! I'm glad people are trying to address the issue of generating SQL within Clojure.)
It's also worth mentioning Pedestal[1] - in place of Ring and Compojure. Pedestal is set of libraries for building web services that is fast and secure by default. It fully embraces async capabilities (unlike Ring which is inherently synchronous), and maintains most Ring compatibility. It sits upon battled-tested technologies (Jetty, Tomcat, JBoss, Undertow, etc).
My team is looking into using Closure for some smaller micro-services and it seems to be a good fit. Forgive my ignorance, but do Clojure applications typically run inside of an application server framework like Tomcat/Glassfish?
It would be good to have a section which compared this to other Clojure database libs. I've only used Korma but I would like to know what you're doing differently to them.
14 comments
[ 3.0 ms ] story [ 37.7 ms ] threadIt believe it says something really good about this approach that we both came to such a similar conclusion independently.
I think OJ takes a slightly more minimalistic approach. I would say that my specification is a little more precise as well. Maybe in the future the author of HoneySQL and I can collaborate.
I'm talking about the kind of stuff something like Korma or SQLAlchemy enable where you build up a query to be executed dynamically based on, for example, user filters.
I haven't need anything like that on any of my pet projects so I've been deliriously happy with Yesql. But, while working on those pet projects I do often find myself thinking about whatever I'm doing in my day job and I haven't come up with a good way to support composability within the context of Yesql. I would love to see a solution to that.
1. Queries are executed by passing your query and db-config to oj/exec. So, does it mean that with any new query there will be new connection to db? How do you manage transaction, when doing inserts/updates? I suppose there must be some magic I don't understand, because if that stuff actually is unsupported that would make all thing pretty much unusable, wouldn't it? Well, if you are not using MySQL/MyISAM anyway.
2. I suppose that library does some internal sanitizing, because it is what such libraries usually are for after all. There isn't much sense in using keywords instead of strings if it is the same plain SQL you would write manually (well, you wouldn't make a typo in the word WHERE, but that's all). But I don't see any mentions of types whatsoever. So if I pass some variable foo to that query, how does it know how to validate and sanitize it? If it is allowed to be NULL? If it should be passed to query as INT or string or maybe DATETIME? As it is minimalistic I would assume it doesn't use schema (and how would it if it's not an ORM, but just some query composer?), so how does it work then?
There are few other things I frowned on, but these are the major points.
1. When you "exec" a query, all I'm really doing is compiling the query map into an SQL statement and passing it into java.jdbc[1] with the database configuration that is given. This database config can be stored as a regular Clojure map in a config namespace in your project or however you wish. To answer your question, yes, at the moment java.jdbc is called for each query that is executed and I can only assume that that means that a new connection is made. Obviously there is room for improvement here, which will be addressed in the future.
2. I've provided a thorough specification for what makes a Clojure map a "query map" in the SPEC file[2]. Before any query map is compiled into SQL, it gets validated by a function that ensures that it meets the specification. If not, it throws an error.
Hopefully this answers some of your concerns! Like I said, this is a new project that I made for fun and for personal use. I plan on working on it for a long time, and I'm sure that many things will be improved! I hope that others find it useful (or at least interesting) as well.
[1]: https://github.com/clojure/java.jdbc/
[2]: https://github.com/taylorlapeyre/oj/blob/master/doc/SPEC
(And btw, your library looks really quite good! I'm glad people are trying to address the issue of generating SQL within Clojure.)
Leiningen [2] is the standard build tool, and Clojars is the standard package repository [3].
For web development, Ring [4] is the standard HTTP solution, while Compojure [5] is generally used for routing.
http://www.clojure-toolbox.com is good for everything else.
[1]: http://www.braveclojure.com
[2]: http://leiningen.org
[3]: https://clojars.org
[4]: https://github.com/ring-clojure/ring
[5]: https://github.com/weavejester/compojure
[1] https://github.com/pedestal/pedestal
My team is looking into using Closure for some smaller micro-services and it seems to be a good fit. Forgive my ignorance, but do Clojure applications typically run inside of an application server framework like Tomcat/Glassfish?
https://github.com/taylorlapeyre/oj/issues/4