A proof of concept tying several potential parts together.
read-kinds (or note-kinds)
A tiny library to produce Kindly advice from notebooks expressed as Clojure namespaces. Kindly advice informs visualization tools how to display code and values. This library is intended for use by literate programming tools such as Clay which produce documents from code.
Clojure namespace -> Kindly advice (what to display)
Pre alpha. Subject to change.
No releases are provided yet. Please add Claykind as a git dependency to your project if you would like to try it out.
Claykind can be invoked from the commandline (see scicloj.claykind.main/-main
)
or the REPL (see scicloj.claykind.api/render!
).
clojure -M:dev -m scicloj.claykind.main --help
Claykind
Version: 0.1.62-alpha
Description: Claykind evaluates Clojure namespaces into Markdown
Options:
-f, --formats FORMATS
-p, --paths PATHS
-t, --target-dir DIR
--evaluator [:clojure|:babashka]
--verbose
--quiet
-h, --help
-v, --version
(require '[scicloj.claykind.api :as claykind])
;; render the entire project
(claykind/render!)
;; render the current file being edited
(claykind/render! "notebooks/test/basic.clj" {:verbose true})
Options can be supplied by a file claykind.edn
in the root of your project,
passed from the commandline, or passed as an argument when calling clay/render
.
Options are the result of merging the defaults, config file, and arguments.
See scicloj.claykind.main/cli-options
for an overview of options available.
By default, Claykind will look for source files in a notebooks
directory (:paths ["notebooks"]
),
and produce output to a docs
directory (:target-dir "docs"
).
See scicloj.claykind.api/default-options
for a full list of default options.
Markdown front-matter is useful when using publishing tools like Quarto. Front-matter will be taken from the metadata of the first form in the notebook.
Source | Compiled |
---|---|
notebooks | docs |
babashka-example-book | _book |
how-to-code-a-book | _book |
A glue library can be made use of in more contexts. The Kindly specs can be used for the creation of markdown, tests, or other purposes.
Clay does quite a lot of things, and has undesirable dependencies. We want to attempt to build something even simpler. Something that emphasizes the evaluation and preparation. Leaving downstream tools to perform rendering and display or test for changes.
Focusing on generating Markdown gives users more flexibility in selecting complementary publishing tools.
The key design focus is on creating interoperable parts, and thus the key concern is what shape should the data be?
Taking a broader view, we want all these parts to come together to achieve several outcomes:
Many of these parts already exist and have existing protocols that we should adapt to.
A complete pipeline of tool composition might look something like this:
my-ns.clj
-> AST
-> top level forms
-> eval
-> kind inference
-> kind validation/parsing
-> downstream tools (editors/html/tests)
Throughout the pipeline will pass various representations, beginning with a string (the original representation of the code), which is read into an Abstract Syntax Tree, which is converted into forms for evaluation.
Let’s examine the stages of the pipeline.
The first stage converts from the text representation of the code into a data-structure. The shape of this data-structure will remain largely unchanged from here, being annotated with more information by other stages.
[{:code "(...)"
:form (...)}
{:code "(...)"
:form (...)}]
You can imagine the above representing all the top-level forms in a namespace.
[{:code "(...)"
:form (...)
:value {:column-names ... :row-maps ...}}
{:code "(...)"
:form (...)
:value {...}}]
[{:form (...)
:code "(...)"
:value {:column-names ... :row-maps}
:kind :kind/table}
{:code "(...)"
:form ()
:value {...}
:kind :kind/vega-lite}]
^:kind/table {:rows [^:kind/table {:rows [1 2]]}
so having a :value
and
a :kind
is insufficient. Either the value should be replaced with a metadata annotated value, or the kind should be
a richer version of the value in some other way. On the other hand, the top level kind may still have some use.[{:form (...)
:code "(...)"
:value {:column-names ... :row-maps}
:kind :kind/table
:valid true
:parsed {...}}
{:code "(...)"
:form ()
:value {...}
:kindly :kind/vega-lite
:valid false
:exception {:message "Bad"
:line 34}}]
Tools benefit from having a standard way of passing information about things to display. We have already seen what that looks like. Kindly defines the tool passing shape (which end users don’t need to know about), and it also defines the user facing grammar of what things can be displayed. End users experience Kindly as a schema for creating data structures that can be visualized.
Kindly abstracts inference. Kindly abstracts rendering. Imagine a blog post without any tool specific annotations, that can generate Markdown, HTML, and be viewed in an IDE.
There is a short list of Kindly grammars:
End users have several options for informing tools which grammar they intend to use:
kind
via metadatakind
via metadata.Advice in itself may not be enough for the visualization tool to work with, as it may not be designed with Kindly in mind. In such cases, plugins exist to translate grammars. See kind-portal as an example.
There are many tools that compose this way:
digraph G {
Code -> Kindly
Kindly -> Clay
Kindly -> Portal
Portal -> Calva
Portal -> Cursive
Portal -> Browser
Kindly -> Markdown
Markdown -> Quarto
Markdown -> Pandoc
Quarto -> HTML
Quarto -> PDF
Quarto -> RevealJS
Pandoc -> HTML
Pandoc -> PDF
}
This library concerns itself only with constructing the initial data representation of the code and pairing it with values from evaluation.
A big challenge in achieving the composition we seek is establishing standards to abide by for tool creators. Perhaps we can draw inspiration from HTML strategies.
Adapters take a context and produce a representation, for example hiccup, markdown, or portal-annotated values.
Builders make pages.
Claykind creates an interface for rendering notebooks to markdown via commandline and code or REPL.
Read-kinds evaluates notebooks into a sequence of contexts.
Copyright © 2023 Scicloj
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.