A friend of mine who is a computer scientist told me he thinks SQL is a thing of the past. I used SQL for years and think the language is totally relevant when dealing with database specific tasks.
My friend believes the new frameworks (e.g.: django, etc) that are coming out are making things even more generic, making SQL an outdated database language. I am wondering what others think about this.
SQL could be better, but it does what it does better than anything else.
SQL has been fine tuned to provide a lot of control for people wanting to perform very specific queries on a database. Even if the databases shift (from RDBMS' to distributed key-value pair dictionaries, for example), there's no reason to think a knowledge of SQL won't be invaluable. I think that SQL, as far as a database language, is what the distributed systems guys will reach for (as a simple key-value pair is inherently limited).
If you are just building crud applications, then it might appear like it is outdated, for more sophisticated queries it would be a pain to do using an ORM, or the code will end up longer than the original sql query.
SQL is a query language for relational databases, it is quite mature and has the functionality that you will need to query relational data.
The frameworks your friend are thinking of are ORM solutions, that try to hide the relational part of the database and mimic it as an object-oriented database, so you can use OO-like fetures to access your data.
This might mean that the future is object-oriented databases, but no one has prooved this claim yet, mostly because relational databases are quite well performing, and ORM tools can hide the relational part of their design.
And of course there are other databases which are emerging, like hiearchical XML databases, with XQuery as the query languge. XQuery 1.1 is quite good in terms of expressiveness, and there are some tools that will let you use XQuery to query relational databases but this won't mean that SQL is outdated.
There are many people who would like to think that it is outdated. There are many reasons for this but basically it comes down to the relational-object mismatch. A lot of effort has gone into creating ORMs ("object relational mapping") that abstract away the database. In short, SQL is being treated as a problem that needs to be fixed.
Unfortunately abstractions are leaky.
Databases and SQL will be around for some time to come.
Jeff Attwood stated on Hanselminutes that he thought that "SQL is like assembly language for the database" and I tend to agree. I also agree that SQL could be better but it gets the job done quite well. And if you work in the Microsoft world then you can do C# in the database directly, bypass sql all together in favor of LINQ to SQL or NHibernate, etc. then you have a lot of options that allow you to move around writing direct sql. But at the end of the day...when I need a really optimized query...you can't beat the underlying SQL! ...sort of the same reason that assembly is still floating around these days!
SQL isn't going anywhere.
There might be other alternatives (e.g., object databases, Big Table, graph databases), and people who don't like SQL (e.g., Anti-SQL), but rest assumed that relational databases will be around for a long time. Your bank will be using relational databases after we're all dead and gone.
Frameworks like Django abstract away the SQL from you, because it can be tedious. They still use SQL and couldn't function without it. It does what it needs to. Does it have flaws, sure, but what language doesn't. , personally, don't like how NULL is neither true nor false.
Now, I almost agree that learning SQL is outdated. There are many ORMs out that take away the need to know it. However, I don't think that a good software engineer can really go without it. I have worked with a number of people who don't know how to make a query, and just sit there dumbfounded. Then when they figure it out, they make queries that are inherently vulnerable to injection.
I think that it's possibly true that relational databases and SQL aren't the perfect fit for web applications. After all, both RDBMs and SQL is considerably older than the web, and were mostly designed for applications with quite different needs than our current. So in that sense I think its true that it is outdated to some extent, and we'll probably see a decline in database use for internet applications.
Lets not forget though that the kind of business applications for which SQL was originally designed is still alive and well, so it's a safe guess that relational databases will be around for a long time.
The fact that your friend did not explain / realize / understand that Django uses SQL for its underpinnings pretty much invalidates the argument he attempts to make.
ORM are a fairly costly abstraction layer when you are looking to deal with high performance applications / sites. If you are doing complicated database queries (multi table unions / intersections etc) and using an ORM it is likely you well send the "raw sql" directly through the ORM to the database.
I don't think that SQL as a concept is out of date - a data access language that describes what is to be done rather than how its to be done is still relevant. The fact that most business systems in production, entering production or on the drawing board that require data storage are still built around SQL says that its still relevant.
What has not kept pace is the adoption of newer elements of the standard and the pace of innovation in the standard lags seriously from demands of modern applications.
Read this: "Why use Stored Procedures?". Ignore the title though: read the content.
Your friend answered his own question: SQL has outlasted all the new frameworks because "data has a long lifespan": frameworks don't.
SQL as a language has evolved over time (SQL-87, SQL-89, SQL-92, SQL:1999, SQL:2003, SQL:2006, SQL:2008) to satisfy ever-growing demand for database technologies in general and relational database technology in particular.
Comparing of SQL to database assembler language is absolutely correct analogy - but it's assembler in reverse: all its higher-level derivatives are in fact lower level languages representing mix of ORM frameworks and less successful alternatives or simplified derivatives.
The only database language that can claim to be higher level than SQL is post-relational database PICK-BASIC et al. But they were never successful in overtaking SQL's place in database world, in part, because of a solid place relational database technology occupies as data platform of choice in majority of software systems of all ranks and levels. And this solid place relational databases almost 100% own to declarative power and expressiveness of SQL.
So there is a self-sustained system comprised of two elements - database technology (relational) and its language (SQL) - and both components propel each other to ever-growing dominance.
Next question should be what would it take to overcome dominance of relational databases and SQL? There were hierarchical databases, object-oriented database, xml databases. After each attempt SQL simply contained new attempts as additional feature set and dismissed technologies as a whole. My guess that until computer power gets to completely new level of performance the alternative to SQL (relational databases) is hardly possible. The new cloud computing technologies are first signs of such alternatives though.
Not only is SQL outdated, it is an anachronism. RoR's ActiveRecord and other Object mappings like Django's are easier to use and more portable than SQL.
By SQL I'm going to assume your friend is talking about relational databases in general, which certainly aren't going anywhere - businesses require them for datawarehousing, and reporting across huge disparate systems. There is a definite trend however toward distributed applications where Key/Pair Databases such as BigTable and CouchDB make a lot of sense. I suspect these newer technologies will become more commonplace on the web, but SQL and relational databases will continue to be used in the Enterprise.