looks like you've been hell-banned, bobbyi [for people who don't have showdead enabled, they suggested hibernate].
my understanding of hibernate is that philosophically hibernate is more like django-orm than sqlalchemy, in that it doesn't embed sql in an dsl (you either map objects directly or use hql in strings). the kind of think i was meaning by "sqlalchemy in java" was something more like c#'s linq, or the scala equivalent(s?).
but i may be wrong, as i haven't used hibernate much. in which case please educate me...
[some context for java people: sqlalchemy will let you define an sql schema in python, then create the database from that, and then let you construct - again, in python (objects and chained method calls) - sql queries against that database. so it's more like "sql in python" than an orm, although it does also support orm if you want (and i suspect because of that sqlalchemy's orm is likely closer to ibatis than hibernate).]
Well DSLs aren't very feasible in Java due to the language, so they came up with HQL instead - the goal is the same, a rapid way to describe SQL statements in terms of objects. In my view, there's not a huge difference between parsing HQL into a tokenized structure, versus building the tokenized structure directly using a generative system like `query(class).filter(class.foo=='bar')`. They do also have a slightly more DSL-like thing called "criteria" queries, which are of the "criteria.add(Restrictions.like("name", "ed"))" variety, a far cry from linq though.
But an ORM is about a lot more than just building SELECT queries. Hibernate has a Session (SQLAlchemy copied this name from Hibernate), built on top of similar unit of work/identity map patterns - it also uses pretty much the same notion of transaction integration (that is, flush() any number of times in a transaction, including before queries, then commit() at the end). SQLAlchemy originally copied the name for their "save_or_update()" method, until taking a cue from Storm and renaming it to "add()". The merge() method in SQLAlchemy is directly from Hibernate, to the point where I read its source to determine its exact behavior in some edge cases.
The notion of "cascades" and the options it accepts are almost identical to Hibernate. There's the fact that collections are represented persistently in memory once loaded, rather than doing a "load on every iteration" kind of pattern that it appears all the other Python ORMs do, as this was the pattern that seemed more familiar to me as well as a lot more efficient in a majority of cases.
Collection loading via joins or a second query that loads for a full set of parent objects at once is pretty much inspired from Hibernate, though I sought to dramatically improve Hibernate's behavior here by getting "joined eager loading" to work totally transparently, including if the parent query had LIMIT/OFFSET on it.
Originally we had also copied this Hibernate concept called "entity name mappings" until realizing it was a terrible idea not at all relevant to Python. The horizontal sharding extension, which I've always viewed as just a proof of concept/experimental thing, was designed pretty much verbatim from Hibernate's similarly little-used API...people were really bugging me for a transparent, horizontal sharding system for awhile, even though they still seem unwise to me, so I figured copying their idea was better than just totally guessing how to do it.
Overall, SQLAlchemy can be seen as a product largely derived from Hibernate in Java and SQLObject in Python, rounded out with various patterns in "Patterns of Enterprise Architecture", and later on from the Storm ORM as well.
my understanding of hibernate is that philosophically hibernate is more like django-orm than sqlalchemy, in that it doesn't embed sql in an dsl (you either map objects directly or use hql in strings). the kind of think i was meaning by "sqlalchemy in java" was something more like c#'s linq, or the scala equivalent(s?).
but i may be wrong, as i haven't used hibernate much. in which case please educate me...
[some context for java people: sqlalchemy will let you define an sql schema in python, then create the database from that, and then let you construct - again, in python (objects and chained method calls) - sql queries against that database. so it's more like "sql in python" than an orm, although it does also support orm if you want (and i suspect because of that sqlalchemy's orm is likely closer to ibatis than hibernate).]