Orchard

A Clojure library designed to provide common functionality for Clojure development tools (e.g. Clojure editor plugins and IDEs).

Right now orchard provides functionality like:

• enhanced apropos
• classpath utils (alternative for java.classpath)
• value inspector
• Java class handling utilities
• Utilities for dealing with metadata
• Namespace utilities
• Fetching ClojureDocs documentation
• Finding function dependencies (other functions invoked by a function) and usages

Why?

Much of the tooling code required to build Clojure editors and smart REPLs is tool-agnostic and should be reused between tools, instead of copied and altered in each and every tool.

Having a common tooling foundation typically means:

• Better foundation (e.g. more functionality, good documentation, etc) with more contributors
• Less work for tool authors as they don't have to reinvent the wheel for every tool
• Happier end users

Design

Orchard is meant to be used to build programmer tooling relying on inspecting the state of a running REPL. REPL-powered tooling has been a core Lisp idea for many years and there are many Clojure libraries in that space (e.g. compliment, tools.trace, sayid, etc).

One thing to keep in mind is that Orchard relies (mostly) on runtime information, not the source code itself. In simple terms - only code that's loaded (evaluated) will be taken under consideration. That's pretty different from the static analysis approach taken by tools for most programming languages where it's not possible to easily inspect the state of running program.

Some other design goals are listed bellow.

No Runtime Dependencies

Orchard is meant to run alongside your application and we can't have a dev tools library interfere with your app right? Dependency collisions are nasty problems which are best solved by making sure there can't be any shared libraries to cause the conflict.

API Optimized for Editors

Code editors can't know what symbols resolve to without consulting a REPL that's why they would typically send a combination of a symbol name and some ns (e.g. the current namespace), so they can be resolved to some var on which an operation would be invoked.

That's why the majority of the functions in Orchard take a combination of a ns and a symbol instead of a var. Probably down the road we'll provide var-friendly versions of most functions as well.

REPL Agnostic

No matter whether you're using nREPL, a socket REPL, or prepl, Orchard has your back. nREPL clients might opt to wrap some of the Orchard functionality in middleware for convenience (as cider-nrepl does), but they can just eval away if they please.

API Documentation

Documentation for the master branch as well as tagged releases are available here.

Usage

orchard requires Clojure 1.10+ and Java 8+.

Note

Java 8 is soft-deprecated in Orchard since version 0.29. Core Orchard funcitonality continues to work on JDK 8, but these following features don't:

• Java sources parsing

Just add orchard as a dependency and start hacking.

[cider/orchard "0.28.0"]

Consult the API documentation to get a better idea about the functionality that's provided.

Dealing with Java sources

Orchard interacts with Java source files (.java files) in several ways:

• Locates the Java source files to enable the "jump to definition" functionality.
  • Also enables "jump to file:line" from the printed stacktrace (a CIDER feature).
• Parses Java sources to extract additional information about Java interop targets (constructors, methods).
  • Allows jumping directly to method definition in the Java source file.
  • Extends the documentation for interop targets with Javadoc comments, exact method argument names.

Currently, Orchard is able to find Java source files in the following places:

• On the classpath.
• In the src.zip archive that comes together with most JDK distributions.
• For clases that come from Maven-downloaded dependencies — in the special -sources.jar artifact that resides next to the main artifact in the ~/.m2 directory. The sources artifact has to be downloaded ahead of time, for example, by enrich-classpath.

If the source file can be located, this is usually enough for basic "jump to source" functionality. For a more precise "jump to definition" and for Javadoc-based documentation, Orcard will attempt to parse the source file.

Configuration options

So far, Orchard follows these options, which can be specified as Java system properties (which means that end users can choose to set them globally without fiddling with tooling internals):

• "-Dorchard.initialize-cache.silent=true" (default: true)
  • if false, the class info cache initialization may print warnings (possibly spurious ones).

Development

Having JDK sources archive ($JAVA_HOME/lib/src.zip) is important for development of Java-related features in Orchard. Certain features parse those Java sources as a source of information. The archive doesn't need to be on the classpath, it just need to exist in the distribution.

You can install Orchard locally like this:

PROJECT_VERSION=99.99 make install

For releasing to Clojars:

git tag -a vX.Y.Z -m "Release X.Y.Z"
git push --tags
git push

Tests and formatting

To run the CI tasks locally use:

make test cljfmt kondo eastwood

Caveats and Known Issues

xref/fn-deps and xref/fn-refs limitations

These functions use a Clojure compiler implementation detail to find references to other function var dependencies.

You can find a more in-depth explanation in this post.

The important implications from this are:

• very fast
• functions marked with meta :inline will not be found (inc, +, ...)
• redefining function vars that include lambdas will still return the dependencies of the old plus the new ones (explanation)
• does not work on AoT compiled functions

Java 8 support

As noted earlier Java 8 is soft-deprecated in Orchard since version 0.29. Core Orchard funcitonality continues to work on JDK 8, but the following features don't:

• Java sources parsing

We are aware that some people are stuck using Java 8 and we'll keep supporting for as long as we can, but it's no longer a priority for us that every feature works with Java 8.

History

Originally SLIME was the most popular way to program in Clojure with Emacs and a lot of useful functionality was created for it to support things like code completion, value inspection, finding references, apropos and so on. This functionality was implemented as a swank adapter written in Clojure and lived in the swank-clojure project.

Subsequently CIDER and cider-nrepl replaced SLIME and swank, and much code was moved from swank-clojure to cider-nrepl and continued to evolve there.

You can watch the presentation The Evolution of the Emacs tooling for Clojure to learn more about all of this.

This project is an effort to prevent repeating the mistakes of the past - cider-nrepl was split into two libraries, so that non-nREPL clients can make of use of the general functionality contained in cider-nrepl (e.g. things like apropos, inspect, etc).

License

Copyright © 2018-2024 Bozhidar Batsov & contributors

Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.