README.md

About

It's a t-sql parser library, mostly entirely mssql flavored, but the scanner is fairly agnostic and I plan to extend it later for other vendors.

The long-term vision for this project is to integrate with my other datbase management project and provide a sql linter / analyzer / code fix provider. Maybe as a SQL Ops studio extension or something.

Status

• scan huge list of sql tokens
• math ops / comparisons / case expressions / function calls
• select / from / where / order by / group by / having
• create
  • table / view / proc / statistics
  • with (options...)
  • contraints inside create-table
  • functions
  • about a billion more types
• update
• delete
• drop
  • table
  • procedure
  • view
  • function
  • database
  • schema
  • index
  • like 100 more dropppable objects
• alter
  • table (20%)
  • procedure
  • view
  • ... more
• in / exists / like / not
• any / all / some (50%)
• Common Table Expressions
• fulltext search
• cursors
• transactions
• Warehouse Features
  • Create Table As Select (buggy)
  • Create remote table as select
• TOOLS
  • AST pretty-printer
  • example linter / analyzer with some decent starter rules
  • dependency graph generator

Shower Thoughts

Format with symbolic names

Pretty-print code based on the underlying table schema.

SELECT * FROM SOME.TABLENAMEHERE
-- to
select * from some.TableNameHere

Vizualize the relationships between scripts with graphviz

node .\dist\src\index.js graph .\test\mssql\kitchen_sink.sql | dot -Tsvg | out-file graph.svg

Module: Data Governance

• [Change Governance] Detect logical conflicts between multiple changes.
• Trace risk impact of a change against an entire system
• Reverse a field to find all the upstream calculations
  • include filters and possibly data volume information?
    • 5M -> {filter} -> 2m -> {filter} -> 100k
• Mass-Refactoring: flow a data type change or column rename through an entire codebase and generate the required
• Experiment: test a change (at a small scale) to see how a code change will affect the end result.

Z3 Server

Read about SAT-solvers on reddit recently.

Currently, I have a naive check for obviously UNSAT conditionals, such as a != a, however I can think of some less obvious cases that could be automatically proven UNSAT, such as len(anything) < 0, (a + b) != (b + a) and things like that.

So, my hair-brained idea is to write some sort of crazy out-of-process z3 server with a grpc binding.

Encode all boolean expressions as theorems and attempt to prove SAT, or declare the code unreachable.