This Relational Algebra evaluator is written by me, Morten Fangel (fangel@sevengoslings.net, twitter.com/fangel). For their assistance with this project, I'd like to thank Assistant Professor, Ph.D. Henrik Bulskov from Roskilde University and Associate Professor Philippe Bonnet, IT University of Copenhagen.
Feel free to contact me if you have any issues or ideas for further work.
RelAlg.js is now in version 0.2.1, and consists of a web-application and a command-line interface. The web-application can be seen at http://fangel.github.io/RelAlg.js, and the CLI can be installed using npm install relalg.js.
RelAlg.js consists of a main library (lib/relalg) that contains a parser, a static type-checker and a evaluation tool. Then a Boostrap/Flight powered web-application (lib/relalg/web) and a Node.js REPL-style CLI (lib/relalg/cli).
All the available functions takes input like this Operation[arguments](Relation). Binary operations are
preformed using Relation1 Operation Relation2.
The REPL is built on top of Node.js, so you can use the special Node.js commands. The two most useful are .save and .load.
.save [filename] will simply take all the operations you have performed in the current session and save them to a file with the name filename.
.load [filename] will read in the file specified by filename and run the commands, one line at a time.
This enables you to have a file that defines a set of operations that you can start all your sessions out by loading them. In the future I will probably add a command-line parameter to a file with initial relations, but till then, the .save and .load will have to do.
You can either reference a relation by it's name (if you have previously assigned a relation), or you can create anonymous relations with the syntax
[['attribute_a', 'attribute_b'] -> [1,2], [2,3], [3,4]]
Projections are done via the Project operation. And example where the columns a and b are projected would be
Project[a,b]( Relation )
Renaming are done with the Rename operation. If you want to rename the attribute alpha into a the command would be
Rename[alpha/a]( Relation )
Selection are done with the Select operation. It takes a boolean expression as its arguments, ie
Select[a == 2 && b <= 3]( Relation )
And and or are expressed with && and ||. The supported comparison operations are:
==(equals)!=(not equals)<=(less than or equals)<(less than)>=(greater than or equals)>(greater than)
Strings are represented quoted in '-s, ie 'this is a string'. If you want to use '-s inside your string, escape them by adding an extra ', ie 'it''s doubles all the way'.
Unions are made using the Union binary operation. Note that the two relations must have the same attributes, NOTE: in the same order!.
Example:
Relation1 Union Relation2
Intersections are calculated with the Intersect binary operations. Note that the two relations must have the same attributes, NOTE: in the same order!.
Example:
Relation1 Intersection Relation2
Set differences are calculated using the - binary operations. Note that the two relations must have the same attributes, NOTE: in the same order!.
Example:
Relation1 - Relation2
The Cartesian product of two relations is created with the X binary operator.
Example:
Relation1 X Relation2
Warning: If the two relations both have an attribute with the same name, there is a naming conflict. This will cause the calculations to fail. You must ensure no duplicate attributes.
Joins are created using the Join binary operation.
If you wish to create a theta-join, you provide the join-condition like this:
Relation1 Join[attribute1 == attribute2] Relation2
Warning: If both relations have attributes with the same name, the calculation will fail. If your intention was to join on the attributes, use a Natural Join. If your intent wasn't to join on those conditions, you must rename the attributes for at least one of the relations.
There is no specific notation for creating Equi-joins, just create a all-AND, all-equality condition and your Theta-join will classify as a Equi-join.
If you do not give a condition as an argument to the Join operator, it will be treated as a natural join.
In other words, natural joins are created like this:
Relation1 Join Relation2
The standard operator for divisions (/) is used to calculate the division between two relations.
An example would be
Relation1 / Relation2
The lexer/parse is built using Jison. So to extend the grammer, you need to have it installed.
When Jison and Grunt is installed, you can rebuild the parser from the grammar-file (src/relalg.jison), using
grunt build.