Hydrilla

# User manual

This page documents basic usage of Hydrilla. The instructions assume a POSIX shell and a UNIX-like system are being used.

{{toc}}

## Installation

### Using Python wheel

You can install Hydrilla server and(or) builder using .whl files from the [[Releases]] page. Please consider [[hachette:Verifying signatures|verifying the downloads]] using provided cryptographic signatures.

#### Installing dependencies

##### Python3

Hydrilla requires Python interpreter in at least version 3.7. You'd typically use Python3 as provided by you operating system distribution. For example, on Debian-based systems (including Trisquel) you can install it with:

``` shell

sudo apt install python3

```

##### pip

Pip is the package manager for Python. While not a direct dependency of Hydrilla, it is needed to utilize .whl files. You most likely also want to install pip as provided by your distro, e.g. for APT-based ones:

``` shell

sudo apt install python3-pip

```

##### Python libraries

Hydrilla relies on the following Python packages:

* `jsonschema`

* `click`

* `flask` (needed for Hydrilla server only)

* `reuse` (optional, only needed for Hydrilla builder to generate SPDX report)

If you don't have those dependencies installed, pip will automatically pull them from [PyPI](https://pypi.org) (except for reuse which would need to be installed separately with a command like `python3 -m pip install reuse`).

Nevertheless, you are encouraged to instead install the respective packages from your operating system's official repositories because those usually have stricter policies on stability, security and free licensing. In case of APT-based distributions the packages to install would be `python3-jsonschema`, `python3-click`, `python3-flask` and `reuse`[^reuse].

[^reuse]: Reuse tool was first packaged for Debian Bookworm and is not yet available in Debian Bullseye nor in Trisquel Nabia.

#### Installing Hydrilla

Let's assume you want to install version 1.0 of Hydrilla server. First, download and verify both[^server_depends_on_builder] `hydrilla.builder-1.0-py3-none-any.whl` and `hydrilla-1.0-py3-none-any.whl`. Then, run:

``` shell

python3 -m pip install path/to/downloaded/hydrilla.builder-1.0-py3-none-any.whl path/to/downloaded/hydrilla-1.0-py3-none-any.whl

```

[^server_depends_on_builder]: Hydrilla server also depends on Hydrilla builder.

This will install Hydrilla **for the current user**. The commands `hydrilla` and `hydrilla-builder` will be made available in `~/.local/bin/`.

#### Installing in virtualenv

If for example you don't want pip to install things under `~/.local/`, you might choose to create a virtual Python environment. First, make sure you have the `virtualenv` tool installed[^virtualenv_tool] (for example from APT package `python3-virtualenv`). Then, choose the folder in which you'd like to install the environment and run:

``` shell

virtualenv -p python3 --system-site-packages path/to/chosen/folder

```

[^virtualenv_tool]: Since Python 3.3 a virtual environment can also be created without this tool.

The `--system-site-packages` flag is not strictly necessary for it to work but is needed if you want packages inside the virtual environment to be able to see globally-installed dependencies.

Once the environment is created, you need to enter it by sourcing a script created by the virtualenv command, e.g.:

``` shell

source path/to/chosen/folder/bin/activate

```

Afterwards, the `python3 -m pip` commands you enter in this shell will install packages inside this virtual environment. You can learn more about Python virtual environments from online tutorials and the [virtualenv documentation](https://virtualenv.pypa.io).

### Using APT

Hydrilla APT repository is hosted at https://hydrillarepos.koszko.org/apt/ and is signed with Wojtek's PGP key (fingerprint **E9727060E3C5637C8A4F4B424BC5221C5A79FD1A**). It is expected to work with modern releases of most APT-based distributions (including Debian bullseye and Trisquel nabia).

This APT repository can be used to install Hydrilla server and builder system-wide and to later update the installation. It has to be said that this also requires you to trust Wojtek's repository with your system's safety (a malicious APT repository could easily take over a system that uses it).

If you've decided you want to install the APT repository on your system, the easiest way to do so is by copy-pasting the following script into your POSIX shell (and then confirming with your password). You can of course modify it according to your needs.

``` shell

__install_hydrilla_apt_repo() {

    local TMP="$1"

    local LISTS="$(cat <<EOF

deb     https://hydrillarepos.koszko.org/apt/ koszko main

deb-src https://hydrillarepos.koszko.org/apt/ koszko main

EOF

)"

    if ! wget -O "$TMP/koszko-keyring.gpg" https://hydrillarepos.koszko.org/apt/koszko-keyring.gpg; then

	echo "Error! Failed to download keyring file!" >&2

	return 1

    elif ! gpg --no-default-keyring --keyring "$TMP/koszko-keyring.gpg" --list-key E9727060E3C5637C8A4F4B424BC5221C5A79FD1A; then

	echo "Error! Invalid keyring file! Someone might be doing something nasty!" >&2

	return 1

    elif ! sudo cp "$TMP/koszko-keyring.gpg" /etc/apt/trusted.gpg.d/; then

	echo "Error!" >&2

	return 1

    elif ! printf %s "$LISTS" | sudo tee /etc/apt/sources.list.d/hydrillarepos.list > /dev/null; then

	echo "Error!" >&2

	return 1

    fi

    sudo apt-get update

}

install_hydrilla_apt_repo() {

    local TMP="$(mktemp -d)"

    __install_hydrilla_apt_repo "$TMP"

    local RESULT="$?"

    rm -r "$TMP"

    return "$RESULT"

}

install_hydrilla_apt_repo

```

This snippet is idempotent (i.e. it can be run multiple times and the effect will be as if it was run once). In addition, it executes `apt-get update` command at the end so that your APT is immediately aware of the new repository and its contents.

After installing the repository you can install Hydrilla builder and server using the following commands:

``` shell

sudo apt install python3-hydrilla.builder

```

``` shell

sudo apt install python3-hydrilla # this alone will also pull the builder as a dependency

```

The packages install their modules under `/usr/lib/python3/dist-packages/` which is seen by Python3 interpreters installed from APT. The `hydrilla` and `hydrilla-builder` commands get placed in `/usr/bin/`.

In addition, the `python3-hydrilla` package also includes sample WSGI script and Apache2 config files for Hydrilla under `/usr/share/doc/python3-hydrilla/examples/`.

## Understanding the concepts

Hydrilla serves Haketilo packages through an HTTP interface, as described in [[Repository API]]. It takes the package files to serve from a specific directory in the system as configured by the administrator. The package files stored in that directory conform to [[Hydrilla on-disk data format]]. Since that format is inconvenient for humans to operate on, there also exists another one - the [[Hydrilla source package format]]. One would typically prepare a Haketilo site resource as a source package and then use the `hydrilla-builder` command to convert it to Hydrilla on-disk format.

Hydrilla builder takes a directory with a source package, processes it and (if no errors are encountered) writes the "built" package files into the requested directory. Somewhat counterintuitively, the "build" does not involve actual compilation of sources nor any similar task (in future versions of Hydrilla all these will be delegated to other software packaging systems like Guix). Rather, the purpose of this step is to save files under the desired names (which involve files' hash sums) and to generate complete JSON definitions of packages being processed.

The serveable directory can be populated by invoking Hydrilla builder multiple times to put the files of different Haketilo packages in it. However, it is also possible to "build" multiple source packages into separate directories and then combine them. It is up to you, as the administrator, to choose how you are going to manage built packages, Hydrilla doesn't impose anything in this regard. Just keep in mind that there is no facility to remove a package from a serveable directory and that brutally deleting one package's files could break the other ones.  

## Running

### With development server

Hydrilla repository software includes a `hydrilla` command that can be used to quickly spawn a local repository server. This is unsuitable for deployment of a publicly visible Hydrilla instance but very suitable for testing of both Hydrilla itself and Haketilo packages being developed. The command has an [associated manpage](https://git.koszko.org/pydrilla/tree/doc/man/man1/hydrilla.1) (also included in the APT package) as well as a `--help` option.

For a sample run, you're going to need a directory with some Haketilo packages. You can clone the [source package example repository](https://git.koszko.org/hydrilla-source-package-example/) and perform something along the lines of:

``` shell

mkdir /tmp/tmprepo/

hydrilla-builder -s path/to/cloned/hydrilla-source-package-example/ -d /tmp/tmprepo/

```

Then comes a typical invocation of `hydrilla` command:

``` shell

hydrilla -m /tmp/tmprepo/ -p 0

```

It causes Haketilo packages from `/tmp/tmprepo/` directory to be served on a random free port on `localhost`. Sample run generated this output:

```

 * Serving Flask app "hydrilla.server" (lazy loading)

 * Environment: production

   WARNING: This is a development server. Do not use it in a production deployment.

   Use a production WSGI server instead.

 * Debug mode: off

 * Running on http://127.0.0.1:46485/ (Press CTRL+C to quit)

```

One could then (in another shell) try running some commands like the ones below to confirm that the just-spawned local server is responding:

``` shell

# The following assume that Hydrilla is loaded with the sample Haketilo package

curl http://127.0.0.1:46485/mapping/helloapple.json

curl http://127.0.0.1:46485/resource/hello-message/2021.11.10

curl http://127.0.0.1:46485/query?url=https://hydrillabugs.koszko.org/a/b/c

```

### With Apache2

This section describes how to configure an Apache2 virtual host to serve a Hydrilla repository. This guide is mostly meant to be useful to people running their own web servers.

You're going to need:

* root access on the machine[^rootless_hosting] (for writing to `/etc/apache2/sites-available/` directory)

* Apache2 with `mod_wsgi` installed and enabled

* Hydrilla installed

[^rootless_hosting]: If you want to run a Hydrilla server on shared hosting without root access, this might be achievable using a .htaccess file but is not documented right now.

First, choose a directory where you want to store your serveable Haketilo packages. The default is `/var/lib/hydrilla/malcontent`. You can override this by saving the following file as `/etc/hydrilla/config.json`:

``` javascript

{

    // Path to directory from which Hydrilla will load packages metadata and serve files.

    "malcontent_dir": "/your/chosen/dir"

}

```

Fill the directory with some package files. You might for example clone the [source package example repository](https://git.koszko.org/hydrilla-source-package-example/) and build it with something along the lines of:

``` shell

sudo hydrilla-builder -s path/to/cloned/hydrilla-source-package-example/ -d /var/lib/hydrilla/malcontent/

```

Once done, grab Hydrilla's [sample WSGI script](https://git.koszko.org/pydrilla/tree/doc/examples/hydrilla.wsgi) and save it in your chosen location (the suggested one is `/var/lib/hydrilla/wsgi/hydrilla.wsgi`). Follow the comments in this script to modify it according to your needs.

Now, get the [sample Apache2 configuration](https://git.koszko.org/pydrilla/tree/doc/examples/hydrilla.example.com.tls.conf) (there is also [one for TLS-less deployment](https://git.koszko.org/pydrilla/tree/doc/examples/hydrilla.example.com.conf)), also modify it according to your needs (in particular, you'll likely want to change `hydrilla.example.com` to some real domain of yours) and save under `/etc/apache2/sites-available/your.chosen.config.name.conf`.

You can now enable the configuration with:

``` shell

sudo a2ensite your.chosen.config.name

```

You also need to reload or restart the Apache daemon for the configuration to be picked up (the command to do that varies between init systems). Once you do so, you can verify that the server is running properly. Consider running something like the following (replacing `hydrilla.example.com` with the domain name you used):

``` shell

# The following assume that Hydrilla is loaded with the sample Haketilo package

curl http://hydrilla.example.com/mapping/helloapple.json

# -v flag will let us verify that the "Content-Type: application/json" header is present

curl -v http://hydrilla.example.com/resource/hello-message/2021.11.10

curl -v http://hydrilla.example.com/query?url=https://hydrillabugs.koszko.org/a/b/c

```

If everything is working as expected (i.e. JSON documents are properly served by Hydrilla&Apache2), you can start populating the "malcontent directory" with built packages of your choice.