Contextual

A two-phase Clojure interpreter. Write an expression once, run it many times with good performance.

Deferred evaluation of Clojure expressions with late bindings of input.

Not A Function

If you read the description above and say, "that's just a function", you're right. So why not a function? Few reasons:

• Dynamic input: creating functions willy-nilly from out-of-process inputs is a potential nightmare.
• Special contexts: By evaluating an expression in a special context, we can use it to represent data templates. Example can be rendering HTML, or even a predetermined HTTP request.
• Symbolic manipulation.
• Limited Metaspace: Clojure functions are compiled to unique class instances. Class metadata is stored in the Metaspace. This space can theoretically run out over the life time of a long running application.

Like An Expression

This library allows the user to store and treat expressions as data, and safely evaluate them in different contexts. Moreover, these expressions can be safely generated based on user input and run inside your application with reasonable performance.

Two Phase Interpreter

• Compile: An expression is compiled to a class hierarchy representing its tree structure
• Invoke: Evaluate the expression with given context. Only method calls, zero interpretations.

Scope

Currently contextual is not a complete Clojure interpreter, but it works, it's fast, and can be used.

Adding support for all of Clojure is on the roadmap, but features which degrade performance might not be added.

Usage

Using contextual involves two phases: compilation and execution.

Execution is always performed via (contextual.core/invoke compiled-expr context-map)

Compilation options:

• lookup: map from symbol to value. Used for resolving symbols during compilation. Can contain any value, from primitive to function, i.e. {'foo clojure.core/println 'bar (->path :x y)} is a valid lookup map.
• symbols registry: This is a map of special symbols to be resolved to constructors for objects which implement the protocols IContext or IStringBuild, to be used as new units of syntax and execution. An extension point for users.

Currently, the following compilations are available:

Expressions

(contextual.core/compile expr), where expr can contain any of the supported symbols or resolvable symbols.

HTTP Requests

(contextual.http/compile request) takes a template of containing they keys url path query-params body form method headers, any of which besides url is optional, and emits an invokable which would emit a map with a corresponding structure after invoking all the expressions contained in it.

Special HTTP options:

• serialize-body: when not false-y indicates the request body should be serialized with the provided body-serializer.
• body-serializer: any function which will serialize the emitted request body.
• serialize-form: when not false-y indicates the request form should be serialized with the provided form-serializer.
• form-serializer: any function which will serialize the emitted request form.
• serialize-query-params: when truth-y will append the query params to the end of the URL instead of emitting them as a map. i.e. {:a 1 :b 2} -> ?a=1&b=2.

Design

Protocols

IContext

• -invoke [this ctx]: Invoke the given object with context ctx. Defaults to identity for Object and nil.

IStringBuild

• -invoke-with-builder [this ctx sb]: Invoke the given object with context ctx and StringBuilder sb.

Records

Contextual uses records to describe behaviors. They behave like their corresponding clojure.core names would, with any difference noted below:

• Map: map container which will -invoke every key and value with context.
• If: Makes branching possible. Will invoke the predicate, then either branch based on the result.
• Fn: function container. Will -invoke all of a function's arguments with context, then apply the function.
• Path: a generic getter for a path of keys in ctx. (path :x :y) will evaluate to whatever value is in path [:x :y] in the context map.
• Or/And.
• Str: Will invoke all its arguments and add their non-nil result to a string builder. Nested Strs won't create intermediary Strings but will use the same StringBuilder.
• Let: Works like you'd expect let to work. Lexical environment is implemented via attached metadata on the context and environment chaining.

Constructors

The defined records aren't meant to be used directly, but are wrapped in lower case constructor functions. An underlying optimization will dispatch to a loop-unrolled record when possible.

Compilation

Given an expression such as

(if (path :x :y)
  (let [x (path :a :b)]
    (+ x 2))
  (str (path :y :z) "blah" (path :u :w)))

Compilation will produce a tree of records representing its structure after a post-walk.

#contextual.core.If{:p #contextual.core.Path2{:k0 :x, :k1 :y}, :t #contextual.core.Let{:bindings #contextual.core.Bindings{:bindings [[x__22910 #contextual.core.Path2{:k0 :a, :k1 :b}]]}, :expr #contextual.core.Fn2{:f #function[clojure.core/+], :a0 #contextual.core.Lookup{:sym x__22910}, :a1 2}}, :e #contextual.core.Str3{:a0 #contextual.core.Path2{:k0 :y, :k1 :z}, :a1 "blah", :a2 #contextual.core.Path2{:k0 :u, :k1 :w}}}

Symbol resolution

Currently, symbols are resolved via:

• symbols-registry
• namespace resolution
• lookup in a map argument

Otherwise, a symbol will be interpreted as an environment lookup.

Performance vs. SCI

Since contextual's model is compile-once run-many, invoke is significantly faster than sci's eval:

(require '[sci.core :as sci])

(def scitx (sci/init {}))

(sci/eval-form scitx '(let [x 1 y 2] (+ x y)))

(def c (-compile '(let [x 1 y 2] (+ x y))))

(-invoke c {})

(require '[criterium.core :as cc])

(cc/quick-bench
 (sci/eval-form scitx '(let [x 1 y 2] (+ x y))))

;;; Evaluation count : 20016 in 6 samples of 3336 calls.
;;;              Execution time mean : 31.412883 µs
;;;     Execution time std-deviation : 1.088819 µs
;;;    Execution time lower quantile : 30.478367 µs ( 2.5%)
;;;    Execution time upper quantile : 33.040048 µs (97.5%)
;;;                    Overhead used : 9.329803 ns

(cc/quick-bench
 (-invoke c {}))

;;; Evaluation count : 543534 in 6 samples of 90589 calls.
;;;              Execution time mean : 1.118617 µs
;;;     Execution time std-deviation : 37.443201 ns
;;;    Execution time lower quantile : 1.088897 µs ( 2.5%)
;;;    Execution time upper quantile : 1.179156 µs (97.5%)
;;;                    Overhead used : 9.414056 ns

In most cases, compiling + invoking contextual code will also be faster than sci.

Status

Experimental, in development

TODO

• [X] Unit tests
• [ ] Tuple records
• [X] Map* records
• [X] Ensure strings work (as advertised)
• [ ] Generalize StringBuilder case to Appendable
• [ ] Check option of similarly implementing OutputStream. Use Writer?
• [X] Bring HTTP request builder up to workable condition.
• [ ] Handle different types of expressions in request better (vector, expr, etc.)
• [X] Faster walk?
• [ ] More macros, (cond!)
• [ ] Improve / control over resolution mechanism
• [ ] Expose only safe functions by default
• [ ] add namespaces
• [ ] Basic interop
• [ ] fns
• [ ] POC tagged template.
• [ ] Replace Records with types
• [ ] Have types' string representation be homoiconic.

License

Copyright © 2020 Ben Sless

This program and the accompanying materials are made available under the terms of the Eclipse Public License 2.0 which is available at http://www.eclipse.org/legal/epl-2.0.

This Source Code may also be made available under the following Secondary Licenses when the conditions for such availability set forth in the Eclipse Public License, v. 2.0 are satisfied: GNU General Public License as published by the Free Software Foundation, either version 2 of the License, or (at your option) any later version, with the GNU Classpath Exception which is available at https://www.gnu.org/software/classpath/license.html.