WORK IN PROGRESS

hreld implements an operator-stream database based on The Third Manifesto (relational algebra).

Tables are text files with simple structure, so they can be managed (created/updated) by ordinary text tools (I am using .tbl extension to easily recognize them).

Columns are separated by a tab character, and rows - by a newline character. The first line in each table should be the header line with column names starting with # character (simplify projection).

During table creation/edition you need to take care of data consistency and duplicated rows removal (we are in the realm of relational algebra).

Examples:

$ cat employees.tbl
#EmpId	#EmpName	#DeptId
1	John Doe	3
2	Jane Roe	1
$ cat former_employees.tbl
#EmpId	#EmpName	#DeptId

• text files inside directories
• /rdb family:
  • NoSQL - with indexes; without row locking. Implemented in AWK.
  • Fsdb - with a rich collection of statistic functions; without indexes and locking. Implemented in Perl.

As any data can be presented in a structured (key-value/tabular) form, you can process any kind of file with simple filters and transformations. The most advanced active projects, that I am aware of are:

• Nushell (CLI) - gets tabular representation of a file (content or metadata) and passes them through UNIX-like pipelined filters;
• Glamorous Toolkit (GUI) - targeting data exploration. You can bind actions to elements of graphic representation. Data processing is easy thanks to well-thought-out methods from Pharo collections.