Programmable, type-safe, human-readable data description language.
PEML is a Haskell embedded domain-specific language (EDSL) that lets you describe your data without visual clutter, using a lean, YAML-like syntax. Just like in YAML and JSON, you build PEML expressions by assembling literal values (strings, integers, etc.) into lists or maps and then nest structures recursively at any depth---yep, it's turtles all the way down. Here's what example 2.3 from the YAML spec looks like in PEML
ex_2_3 = do -- 'do' notation to define expressions
"american" =: do -- the tag operator (=:) names an expression
(-:) "Boston Red Sox" -- whereas (-:) introduces a list item
(-:) "Detroit Tigers"
(-:) "New York Yankees"
"national" =: do
(-:) "New York Mets"
(-:) "Chicago Cubs"
(-:) "Atlanta Braves"We support converting PEML to YAML and JSON, so you can easily dump a PEML expression to a file as YAML or JSON. What the heck, why not write plain YAML or JSON then?! Well, there are a couple of reasons that might make PEML worth your while:
- Type safety. PEML piggybacks on Haskell's advanced type system to make sure you can only put together expressions that translate to valid YAML/JSON. As an added bonus, you can stop worrying about subtle white-space/indentation issues and concentrate on describing your data instead. But you could push the envelope and, e.g. encode your config rules in the type system so you get to know about any blunder through compiler errors rather than a prod outage. Likewise you get the freedom to refactor your config mercilessly as the compiler safely guides you through the process.
- Programmability. After having had truckloads of YAML gently shovelled down your throat with a dozer, you start wondering how to abstract away bits and pieces that look similar but aren't exactly the same. Or how to factor out common patterns in a lib to reuse across projects. Or how to modularise stuff to minimise the impact of change. Or... Ah! You need "programmable YAML"! While code generators and template languages may be a good fit for your use case, they typically clip your wings: you can only write a subset of all possible programs that would be useful. Reaching for time-honoured C&P tech, scripts and all that to get around roadblocks usually leaves you with a hodgepodge nobody dares even looking at. Why struggle? A PEML expression is Haskell data living in a Haskell program which means you can do pretty much whatever you like with it---from defining your own functions to fancy AST transformations, it's all at your fingertips. Also Haskell is a real productivity booster with a best-in-class approach to abstraction, modularity, reuse, and conciseness and comes with a ton of top-notch libs to help you get where you want quick-quick, chop-chop.