jweissman / aeql

query lang for humans

AEQL

a tiny query language for humans

Synopsis

• Define relational models and querying them
• Define your schema declaratively in yaml files
  • Human beings/roles in personae.yml
  • Domain models in models.yml
  • Queries/views/logic in algebrae.yml
• Query language is focused on
  • Simplicity
    • Favor user-defined 'attributes' to narrow resources over complicated things: find most_liked post where most_liked is defined in algebrae?)
    • But you should also be able to do simple computations: find companies where employees.count > average(Company.employee_count)
  • Composability
  • Naturality: find people in toronto whose job is accountant

Ideas

we think that a relational-algebraic grammar which exposes the raw power of that formalism would be great but is kind of untenable

        we think instead we can opt to favor composable of basic operations, and have things spelled <output>
        
        we think tooling that gives lots of feedback early on in the query development process is good

        We think that having a strong opinion about query grammar is important -- we feel like SQL suffers from ambiguity which leads to vendor fragmentation (this is also arguably a strength of the 'ecosystem' but that feels like a problem too)            </output>
          Okay, so a "fresh start" kind of query language!

          Based on human values, and declarative specification of structures

          So you'll give your schema to aeql in the form of a handful of yaml files,
          most notably the `personae.yml` in which you specify the roles of PEOPLE

          The idea is that queries are 'based' around people in general (or the agents,
          systems, etc in your relations: although the idea is that personae are truly
          people and not systems, that we're creating value for people. a simple 'user'
          of the system resources in question...)

          Maybe you don't have to start from people, but the idea is that there's affordances
          and 'naturalisms' we can offer around personae that would highlight the value of doing
          this (of doing user-centered domain modelling)

More ideas

• maybe if left undefined, people matches across ALL personae, even if they don't share a parent model