Develop Reference

Using Mongo recommended to store data inside an hashmap

I just recently switch from MySQL to MongoDB, I'm wondering with MySQL I stored the player data inside a hashmap and retrieved name, coins etc; like that so I don't have to constantly query the database to retrieve the data.
Now with MongoDB would I need to do the same thing store the values inside a hashmap and retrieve it the same way I did with MySQL?

It depends on your requirement. You have migrated to mongodb from mysql, this doesnt means that your reads would be superfast. If there would have been any significant I/O improvement in mongodb, mysql developers would have adopted it as well. MongoDB provide flexibility over mysql and there are some more advantages there. So If your load remains the same, you should have a caching layer before mongodb layer. Both Mysql and mongodb come with in-built caching which caches results on the basis of query just like a hashmap, but rest data is on disk and as mentioned mongodb doesnt have any technical advantage over mysql in terms of I/O. So have a caching layer to avoid excessive querying to db.

How join a record set that is returned from a web service with one of your sql tables

I thought about this solution: get data from web service, insert into table and then join with other table, but it will affect perfomance and, also, after this I must delete all that data.
Are there other ways to do this?

You don't return a record set from a web service. HTTP knows nothing about your database or result sets.
HTTP requests and responses are strings. You'll have to parse out the data, turn it into queries, and manipulate it.
Performance depends a great deal on things like having proper indexes on columns in WHERE clauses, the nature of the queries, and a lot of details that you don't provide here.
This sounds like a classic case of "client versus server". Why don't you write a stored procedure that does all that work on the database server? You are describing a lot of work to bring a chunk of data to the middle tier, manipulate it, put it back, and then delete it? I'd figure out how to have the database do it if I could.

no, you don't need save anything into database, there's a number of ways to convert XML to table without saving it into database
for example in Oracle database you can use XMLTable/XMLType/XQuery/dbms_xml
to convert xml result from webservice into table and then use it in your queries
for example:
if you use Oracle 12c you can use JSON_QUERY: Oracle 12С JSON
XMLTable: oracle-xmltable-tutorial
this week discussion about converting xml into table data

It is common to think about applications having a three-tier structure: user interface, "business logic"/middleware, and backend data management. The idea of pulling records from a web service and (temporarily) inserting them into a table in your SQL database has some advantages, as the "join" you wish to perform can be quickly implemented in SQL.
Oracle (as other SQL DBMS) features temporary tables which are optimized for just such tasks.
However this might not be the best approach given your concerns about performance. It's a guess that your "middleware" layer is written in Java, given the tags placed on the Question, and the lack of any explicit description suggests you may be attempting a two-tier design, where user interface programs connect directly with the backend data management resources.
Given your apparent investment in Oracle products, you might find it worthwhile to incorporate Oracle Middleware elements in your design. In particular Oracle Fusion Middleware promises to enable "data integration" between web services and databases.

Database independency using Hibernate

I am using Hibernate for ORM in my Java application. I want to write custom queries combining multiple tables and using DB functions like sum(salary).
I also want to support multiple databases without writing the SQLs again and again for each database. The approach currently followed
is having Stored Procedures specific to each DB (Oracle, MySQL etc) and whichever we want to support, we change the configuration file in the application.
What I am looking for is a solution very generic so that I need not write Stored Procedures or SQLs for every new functionality.

If you really want to keep it portable, you'll need to do it all with HQL.
There's no reason that you couldn't do multi-table joins and aggregate functions in HQL, you just need to limit yourself to the ones it supports.
Once you start doing database-vendor specific things, you are no longer database independent, by definition.

A perfect suite is HIbernate Criterias
Hibernate provides alternate ways of manipulating objects and in turn data available in RDBMS tables. One of the methods is Criteria API which allows you to build up a criteria query object programmatically where you can apply filtration rules and logical conditions.
http://www.tutorialspoint.com/hibernate/hibernate_criteria_queries.htm

unifying query language for accessing data from heterogeneous database

In my current project, I am trying to unify query language for accessing heterogeneous database. Heterogeneous database means their query language for accessing data is different. For instance, SQL is a query language for accessing data from Apache Derby, while nonSQL for MongoDB.
My question is "Is there any domain specific language, which have been proposed to unify heterogeneous databases ? "
Please feel free to direct me other efforts as well.

That's quite an interesting question. There is at least one proposed solution called UnQL (Unstructured Data Query Language) - http://www.couchbase.com/press-releases/unql-query-language.
I suppose out of the box UnQL will work at least for CouchDB and SQLite. This just seems to be a great step ahead.
Personally I would say such a task seems to be a tricky one because of the conceptual differences between structured and unstructured data approaches. Anyway, it should be relatively easy to develop such a DSL for a well defined SQL and NoSQL data models used by a certain application.

There is a project called Hibernate OGM, which aims to generalize JPQL to NoSQL databases.
From their web page:
Hibernate Object/Grid Mapper (OGM) aims at providing Java Persistence (JPA) support for NoSQL solutions. It reuses Hibernate Core's engine but persists entities into a NoSQL data store instead of a relational database. It reuses the Java Persistence Query Language (JP-QL) to search their data.
I don't tried it out for myself, so I cannot say how usefull it is.

JSONiq can process data from different SQL and NoSQL products.
The open source implementation of JSONiq has connectors for Couchbase, Oracle NoSQL, SQLite, and JDBC.
For instance, the following slidedeck showcase the same query being executed on both Couchbase and MongoDB: https://speakerdeck.com/wcandillon/jsoniq-the-sql-of-nosql

SPARQL is a W3C-standardized query language that works on top of an abstract data model (RDF), rather than a specific type of database, which makes it very suitable as an enabler for heterogeneous database querying.
Implementations of SPARQL exist on top of various NoSQL databases, including native RDF databases (often referred to as triplestores), as well as on top of relational databases.

Does a simple document-based database exist?

Is there a database out there that I can use for a really basic project that stores the schema in terms of documents representing an individual database table?
For example, if I have a schema made up of 5 tables (one, two, three, four and five), then the database would be made up of 5 documents in some sort of "simple" encoding (e.g. json, xml etc)
I'm writing a Java based app so I would need it to have a JDBC driver for this sort of database if one exists.

CouchDB and you can use it with java
dbslayer is also light weight with MySQL adapter. I guess, this will make life a little easy.

I haven't used it for a bit, but HyperSQL has worked well in the past, and it's quite quick to set up:
"... offers a small, fast multithreaded and transactional database engine which offers in-memory and disk-based tables and supports embedded and server modes."

CouchDB works well (#zengr). You may also want to look at MongoDB.
Comparing Mongo DB and Couch DB
Java Tutorial - MongoDB

Also check http://jackrabbit.apache.org/ , not quite a DB but should also work.