This is a tool to evaluate or export code from Markdown files.
Why? Because I like writing Markdown files with code snippets (it's good with Obsidian too).
With smaller code snippets and documentation in a markdown file, the file will
be easily consumed by the reader, yet it can be executed without copy-paste. If
the file has more than one code block with given language, eval-md
will
combine them and evaluate as one script.
cargo install eval-md
- JavaScript (node, deno)
- Lua
- PHP
- Python3
- Ruby
- Shell (bash, zsh)
With the --export
flag, target language can be anything, it will not evaluate
the final code. It will print out the content to stdout
. Ideal to generate
content from a Markdown file, for example configuration files. In the example we
have a JSON configuration file for a service, and we can add extra comments
about sections of the configuration file.
In the output, a header will be added, for example Python scripts get
#!/usr/bin/env python3
on --export
.
Custom tag and executor can be defined with :
. The first part will be the
string tag to extract code blocks, and the second part will be the language that
will evaluate the extracted code.
py:python
=> will parsepy
and run aspython
js
=> will parsejs
and run asjs
(js
is an alias to javascript)js:deno
=> will parsejs
and run withdeno
The following examples will use example/test.md
:
❯ eval-md zsh example/test.md -- --random-flag
nice in zsh
Arguments: --random-flag
❯ eval-md bash example/test.md
nice in bash
Arguments:
❯ eval-md python example/test.md -- --hype-level=awesomeness
awesome
Arguments: ['-', '--hype-level=awesomeness']
❯ eval-md ruby example/test.md -- --debug
it works :)
Arguments: ["--debug"]
❯ eval-md --debug python example/test.md -- --hype-level=awesomeness
-- Target Language: python
-- Source file: example/test.md
-- Arguments: ["--hype-level=awesomeness"]
awesome
Arguments: ['-', '--hype-level=awesomeness']
❯ eval-md json example/test.md --export
{
"enable_registration": true,
"debug": false,
"hostname": "efertone.me",
"port": 9999
}
❯ eval-md lua example/test.md
Value: 15
❯ eval-md something:javascript example/test.md
Fancy
Obviously the whole documentation can live in a shell script as comment. In case you want to add images and links to other pages, you can still "execute" the documentation about how to install Flux. Of course this example installs and bootstraps only one specific way.
❯ eval-md bash example/install-flux.md
!! git repository is not defined.
--help This help message
--repo <repo> Git repository. (example: git@github.com/org/repo)
--branch <branch> Git branch. [default: main]
--path <path> Path to the cluster. [default: ./clusters/management]
❯ eval-md bash example/install-flux.md -- --repo git@github.com/yitsushi/cluster-conf --branch production
...
The example/flux-helm-repo.md
example shows a simple scenario where we
describe how to deploy helm charts on a cluster. Usually that comes with a lot
of copy-paste, but we can be smart with eval-md
.
This simple command will evaluate all bash
in the markdown file. The only
bash
snippet is the one that exports yaml
blocks and pass it to kubectl
.
At the end, we have a nice documentation (not this one, it's mostly just filler
text, did not really spend much time on it).
❯ eval-md bash example/flux-helm-repo.md
helmrepository.source.toolkit.fluxcd.io/podinfo created
helmrelease.helm.toolkit.fluxcd.io/podinfo created
With a special marker, key-value pairs can be defined on code blocks:
```bash #group=a something=else
echo "This is group A"
```
```bash #group=b
echo "This is group B"
```
```bash
echo "This one does not belon anywhere"
```
Without any extra arguments, all bash
blocks will be evaluated:
❯ eval-md bash a.md
This is group A
This is group B
This one does not belon anywhere
If we want to evaluate only a specific group, we can set the filter with the
--group
flag:
❯ eval-md bash a.md --group=a
This is group A
❯ eval-md bash a.md --group=b
This is group B
If --group
flag is specified, but with empty value, it means evaluate
everything that has no group.
❯ ./target/release/eval-md bash a.md --group=
This one does not belon anywhere
With the --pick
flag, we'll be asked about each group to add it to the script
or discard it. This is good if we have more than one code blocks, but we don't
want all of them to be evaluated.
Group selection and Pick mode works together, so it is possible to filter on a group and discard parts from it before evaluation.
Pick mode questions are printed to stderr
, the program output can be
redirected to a file or pipe, it is useful with --export
❯ eval-md bash a.md --group=a --pick --export > myscript.sh
---
echo "This is group A"
---
--> Do you want to add this block? (yes/no) yes
---
echo "This is group A again"
---
--> Do you want to add this block? (yes/no) no
❯ cat myscript.sh
#!/usr/bin/env bash
echo "This is group A"
This is a weird case, but someone said they would use it and that would be cool.
If the name is set to all
with a specified interpreter like all:python
, then
all code blocks will be used in the document regardless the language marker on
them, and it will be evaluated with the specified interpreter (in this case
python
).
See the example/use-all.md
example:
❯ eval-md all:python example/use-all.md
Do something with:
<html>
<head><title>Nice</title></head>
<body>
<div id="app">awesome</div>
</body>
</html>