2colours / collection

Collects content files written in Pod6 (Rakudoc) into (eg) a website

Raku Collection Module

Description Software to collect content files written in Rakudoc (aka POD6) and render them in a chosen format. Extensive use is made of plugins to customise the rendering. A distinction is made between the Rakudoc files for the main content (sources) and the Rakudoc files that describe the whole collection of sources (mode-sources), eg. the landing page (index.html) of a website, or the Contents page of the same sources in book form. The collection process is in stages at the start of which plugin callables (Raku programs) can be added that transform intermediate data or add templates, or add new Pod::Blocks for the rendering.

Author Richard Hainsworth aka finanalyst

---

Table of Contents

Installation
Usage
Life cycle of processing
Modes
Milestones
Stopping or dumping information at milestones
Source Milestone
Mode Milestone
Setup Milestone
Render Milestone
Compilation Milestone
Transfer Milestone
Report Milestone
Completion Milestone
Collection Structure
Collection Content
Extra assets (images, videos, etc)
Cache
Mode
Templates
Configuration
Top level configuration
Mode-level configuration
Plugin level configuration
Control flags
Plugin management
Disabling a plugin
Plugin types
Setup
Render
Compilation
Transfer
Report
Completion
Plugin updating
Mapping released plugins to mode directories
Released plugins directory
Refresh process
CLI Plugin Management System
Problems and TODO items
Archiving and Minor Changes
Dump file formatting
Post-cache methods
multi method add(Str $fn, Array $p --> Pod::From::Cache )
multi method mask(Str $fn --> Pod::From::Cache)
multi method add-alias(Str $fn, Str :alias! --> Pod::From::Cache)
method behind-alias(Str $fn --> Str )
method pod(Str $fn)
Asset-cache methods
Copyright and License

---

This module is used by the module Collection-Raku-Documentation, but is intended to be more general, such as building a personal site.

Installation

To install the distribution, a refresh utility (see [Plugin refreshing](Plugin refreshing.md)), and the default plugin directory (see [Released plugin directory](Released plugin directory.md)), use

zef install Collection

For those who really want to have a non-default plugin directory, it is possible. But warning: extra user input will be needed for other utilities, so read the whole of this file), eg, to have a hidden directory .Collection in the home directory (under *nix), use

PluginPath=~/.Collection zef install Collection

Usage

The Collection module expects there to be a config.raku file in the root of the collection, which provides information about how to obtain the content (Pod6/rakudoc> sources, a default Mode to render and output the collection. All the configuration, template, and plugin files described below are Raku programs that evaluate to a Hash. They are described in the documentation for the RakuConfig module.

A concrete example of Collection is the Collection-Raku-Documentation (CRD) module. CRD contains a large number of plugins (see below). Some plugin examples are constructed for the extended tests. Since the test examples files are deleted by the final test, try:

NoDelete=1 prove6 -I. xt

and then look at eg., xt/test-dir.

The main subroutine is collect. It requires a file config.raku to be in the $CWD (current working directory). In CRD the executable Raku-Doc initiates the collection by setting up sources and installing a config.raku file. It is then simply a command line interface to collect.

Life cycle of processing

The content files are processed in several stages separated by milestones. At each milestone, intermediary data can be processed using plugin callables, the data after the plugin callables can be dumped, or the processed halted.

collect can be called with option flags, which have the same effect as configuration options. The run-time values of the [Control flags](Control flags.md) take precedence over the configuration options.

collect should be called with a Mode. A Mode is the name of a set of configuration files, templates, and plugins that control the way the source files are processed and rendered. The main configuration file must contain a key called mode, which defines the default mode that collect uses if called with no explicit mode, so if collect is called without a Mode, the default will be used.

Modes

A Mode:

• is the name of the process by which a Collection is rendered and presented. At present, only Website is implemented, but it is planned to have Modes for epub and pdf formats. The presentation may be serving HTML files locally, or the creation of a single epub file for publishing.

• is the name of a sub-directory under the Collection directory.

• may not be named with any '_' in it. This allows for sub-directories in a Collection folder that are not modes, eg., a Raku documentation repository

• A Mode sub-directory must contain:

  • a sub-directory configs, and/or a config.raku file.

Milestones

The collect sub can be called once the collection directory contains a config.raku, which in turn contains the location of a directory of rakudoc source files, which must contain recursively at least one source.

Processing occurs during a stage named by the milestone which starts it. Each stage is affected by a set of [Control flags](Control flags.md). Certain flags will be passed to the underlying objects, eg. RakuConfig and ProcessedPod (see Raku::Pod::Render).

Plugin callables may be called at each milestone (except 'Source' and 'Mode', where they are not defined). Plugins are described in more detail in [Plugin management](Plugin management.md). plugin callables are milestone specific, with the call parameters and return values depending on the milestone.

The milestones are:

• Source

• Mode

• Setup

• Render

• Compilation

• Transfer

• Report

• Completion

Stopping or dumping information at milestones

Intermediate data can be dumped at the milestone without stopping the processing, eg.,

collect(:dump-at<source render>);

Alternatively, the processing can be stopped and intermediate data inspected, EITHER after the stage has run, but before the plugin callables for the next stage have been triggered, eg.,

my $rv = collect(:after<setup>);

OR after the previous stage has run and after the plugin callables for the milestone have been triggered, eg.,

my $rv = collect(:before<render>);

The return value $rv is an array of the objects provided to plugin callables at that milestone, and an array of the plugin callables triggered (note the plugin callables used will be a difference between the :before and :after stop points). The plugins-used array is not provided to all plugin callables, except at the Report milestone.

The return value $rv at :after will contain the object provided by the milestone after the named milestone. For example, milestone milestone 'Setup' is followed by milestone 'Render'. The return object for :after<setup> will be the return object for milestone 'Render'. See Milestones for more information.

The object returned by :before<render> may be affected by the plugin callables that are triggered before the named stage.

The :before, :after and :dump-at option values are the (case-insensitive) name(s) of the inspection point for the milestone. :before and :after only take one name, but :dump-at may take one or all of them in a space-delimited unordered list.

The dump-at option calls .raku [TODO pretty-dump, when it handles BagHash and classes] on the same objects as above and then outputs them to a file(s) called dump-at-<milestone name>.txt.

Source Milestone

(Skipped if the :without-processing flag is True)

Since this is the start of the processing, no plugins are defined as there are no objects for them to operate on.

The config.raku file must exist and must contain a minimum set of keys. It may optionally contain keys for the control flags that control the stage, see below. The intent is to keep the options for the root configuration file as small as possible and only refer to the source files. Most other options are configured by the Mode.

During the subsequent Source stage, the source files in the collection are brought in, if the collection has not been fully initiated, using the source-obtain configaturation list. Alternatively, any updates are brought in using the source-refresh list. Commonly, sources will be in a git repository, which has separate commands for clone and pull. If the source-obtain and source-refresh options are not given (for example during a test), no changes will be made to the source directory.

Any changes to the source files are cached by default.

The control flags for this stage are:

• no-refresh (default False)

Prevents source file updates from being brought in.

• recompile

Forces all the source files to be recompiled into the cache.

Mode Milestone

(Skipped if the :without-processing flag is True)

Collection makes a distinction between Rakudoc source files that are the main content, and the source files needed to integrate the main content into a whole. The integration sources will differ according to the final output. For example, a book may have a Foreward, a Contents, a Glossary, etc, whilst a website will have a landing page (eg., index.html), and perhaps other index pages for subsections. A book may also organise content into sections that depend on metadata in the source files. A book will have a defined order of sections, but a website has no order. A website will require CSS files and perhaps jQuery scripts to be associated with Blocks. A book will have different formating requirements for pages.

These differences are contained in the mode configuration, and the plugins and templates for the mode.

At this milestone, the source files have been cached. The mode sub-directory has not been tested, and the configuration for the mode has not been used. Since plugin management is dependent on the mode configuration, no plugin callables can be called.

The return value of collect with :after<source> is a single Pod::From::Cache object that does a Post-Cache role (see below for Post-Cache methods).

A Pod::From::Cache object provides a list of updated files, and a full set of source files. It will provide a list of Pod::Blocks contained in each content files, using the filename as the key.

During the stage the source files for the Mode are obtained, compiled and cached. The process is controlled by the same options as the Source stage. For example, the Mode for Collection-Raku-Documentation is Website.

If a sub-directory with the same name as mode does not exist, or there are no config files in the <mode>/config directory, collect will throw an X::Collection::NoMode exception at the start of the stage.

Mode source files are stored under the mode sub-directory and cached there. If the mode source files are stored remotely and updated independently of the collection, then the mode-obtain and mode-refresh keys are used.

Setup Milestone

(Skipped if the :without-processing flag is True)

If setup plugin callables are defined and in the mode's plugins-required list, then the cache objects for the sources and the mode's sources (and the full-render value) are passed to the program defined by the plugin's setup key.

The purpose of this milestone is to allow for content files to be pre-processed, perhaps to creates several sub-files from one big file, or to combine files in some way, or to gather information for a search algorithm.

During the setup stage,

• the ProcessedPod object is prepared,

• templates specified in the templates directory are added

• the key mode-name is added to the ProcessedPod object's plugin-data area and given the value of the mode.

The Setup stage depends on the following options:

• full-render

By default, only files that are changed are re-rendered, which includes an assumption that if any source file is changed, then all the mode sources must be re-rendered as well. (See the Problems section below for a caveat.)

When full-render is True, the output directory is emptied of content, forcing all files to be rendered.

full-render may be combined with no-refresh, for example when templates or plugins are changed and the aim is to see what effect they have on exactly the same sources. In such a case, the cache will not be changed, but the cache object will not contain any files generated by setup plugin callables.

Render Milestone

(Skipped if the :without-processing flag is True)

At this milestone render plugins are supplied to the ProcessedPod object. New Pod::Blocks can be defined, and the templates associated with them can be created.

The source files (by default only those that have been changed) are rendered.

The stage is controlled by the same options as Setup and

• with-only - affects which Documents are rendered, see Configuration for more

• ignore - prevents docs from being Cached, see Configuration for more

Compilation Milestone

(Skipped if the :without-processing flag is True)

During the stage after this milestone, the structure documents are rendered. They can have Pod-blocks which use data included by templates and plugins during the render stage. They can also add to data, which means that the order in which a plugin is called may be important.

At this milestone plugin callables are provided to add compiled data to the ProcessedPod object, so that the sources in the mode's directory can function.

During the Render stage, the %processed hash is constructed whose keys are the filenames of the output files, and whose values are a hash of the page components of each page.

The compilation plugin callables could, eg, collect page component data (eg., Table of Contents, Glossaries, Footnotes), and write them into the ProcessedPod object separately so there is a TOC, Glossary, etc structure whose keys are filenames.

The return value of collect at the inspection point is a list of ProcessedPod, %process, with the ProcessedPod already changed by the compilation plugin callables.

The stage is controlled by the same options as Setup and

• with-only - same as include but for structure documents

• ignore-mode - as for ignore above

Transfer Milestone

(Skipped if the :without-processing flag is True)

Plugins may generate assets that are not transfered by them, or it is important to ensure that a plugin runs after all other plugins.

In addition, render plugins may create files that are transfered at the render stage, but should be removed after all plugins have run. So a transfer milestone plugin can be created to clean up the plugin's local directory.

Report Milestone

(Skipped if the :without-processing flag is True)

Once a collection has been rendered, all the links between files, and to outside targets can be subjected to tests. It is also possible to subject all the rendered files to tests. This is accomplished using report plugin callables.

In addition, all the plugin callables that have been used at each stage (except for the Report stage itself) are listed. The aim is to provide information for debugging.

The report stage is intended for testing the outputs and producing reports on the tests.

Completion Milestone

Once the collection has been tested, it can be activated. For example, a collection could be processed into a book, or a Cro App run that makes the rendered files available on a browser. This is done using completion plugin callables.

The without-completion option allows for the completion phase to be skipped.

Setting without-processing to True and without-completion to True should have no effect unless

• there are no caches, which will be the case the first time collect is run

• the destination directory is missing, which will be the case the first time collect is run

Collection Structure

A distribution contains content files, which may be updated on a regular basis, a cache, templates, extra assets referenced in a content file (such as images), and one or more modes.

Collection Content

The content of the distribution is contained in rakudoc files. In addition to the source files, there are Collection content files which express things like the Table of Contents for the whole collection.

Collection content are held separately to the source content, so that each mode may have different pages.

This allows for active search pages for a Website, not needed for an epub, or publisher data for an output formation that will be printed.

Extra assets (images, videos, etc)

Assets such as images, which are directly referenced in content file, but exist in different formats, eg, png, are held apart from content Pod6 files, but are processed with content files.

The reasoning for this design is that Pod6 files are compiled and cached in a manner that does not suit image files. But when an image file is processed for inclusion in a content file, the image may need to be processed by the template (eg., image effects specified in a Pod Block config).

The assets are all held in the same directory, specified by the configuration key asset-basenamme, but each asset may exist in subdirectories for each type of asset, specified by the asset-paths key.

(Asset files relating to the rendering of a content file, such as css, javascript, etc, are managed by plugins, see below for more on plugins.)

A class to manage asset files is added to the ProcessedPod object with a role, so the assets can be manipulated by plugins and templates. Assets that are in fact used by a Pod content file are marked as used. The aim of this functionality is to allow for report-stage plugin callables to detect whether all images have been used.

Plugins can also transform the assets, and create new files in the ProcessedPod object for inclusion in the output.

At the end of the compilation stage, all the assets that have been used are written to a directory specified in the Mode configuration file. It is the task of the template rendering block to ensure that the path where the asset is stored is the same as the path the final output (eg. the browser rendering html files) processor requests.

In keeping with the principle that collection level meta data is kept in the top-level config file, and output data is associated with the specific mode, there are two asset-path definitions.

• Collection level assets. The source of assets is kept in the top-level config.raku file. In order to have each asset in its own directory, the following is possible:

...
:asset-basename<assets>,
asset-paths => %(
    image => %(
        :directory<images>,
        :extensions<png jpeg jpeg svg>,
    ),
    video-clips => %(
        :directory<videos>,
        :extensions<mp4 webm>,
    ),
),
...

Notice that the type, eg. image and video-clips above, are arbitrary and not dependent on the actual format.

• Output configuration. The output destination is kept in the mode configuration, eg., Website/configs/03-images.raku contains

%(
    :asset-out-path<html/assets>
    ),
)

For more see [Asset-cache methods](Asset-cache methods.md)

Cache

The cache is a Precomp structure into which the content files are pre-preprocessed.

Mode

The Mode is the collection of templates and configuration for some output. A collection may contain multiple Modes, each in their own subdirectory.

The default Mode for Collection-Raku-Documentation is Website, for example.

The string defining mode must refer to an immediate directory of the root of the collection, so it is compared to / ^ \W+ (\w+) '/'? .* $ / and only the inner \w chars are used.

The templates, configuration, output files, and other assets used by a Mode are associated with the Mode, and should reside beneath the Mode sub-directory.

Templates

The templates, which may be any format (currently RakuClosure or Mustache) accepted by ProcessedPod, define how the following are expressed in the output:

• the elements of the content files, eg. paragraphs, headers

• the files as entities, eg, whether as single files, or chapters of a book

• the collective components of content files, viz, Table of Contents, footnotes, Glossary, Meta data

• All the templates may be in one file, or distributed between files.

  • If there are no templates in the directory, the default files in ProcessedPod are used.

  • If there are multiple files in the directory, they will all be evaluated in alphanumeric order. Note that existing keys will be over-written if they exist in later templates. This is not the same behaviour as for Configuration files.

Configuration

There are three levels of configuration:

• The top-level configuration resides in config.raku in the root directory of the Collection. The collect sub will fail without this file.

• The Mode configuration typically resides in the configs directory, in several files (the names are not important).

• Each plugin has its own config, in the config.raku file of its directory.

Top level configuration

In the descriptions below, simple illustrative names are given to files with configuration, templates, callables. These files are generally Raku programs, which are compiled and run. They will almost certainly contain errors during development and the Rakudo compiler will provide information based on the filename. So it is good practice to name the files that make them easier to locate, such as prefixing them with the plugin name.

config.raku must contain the following keys:

• the cache directory, relative to the root directory of the collection

  • Collection-Raku-Documentation default: 'doc-cache',

• the sources directory, relative to the root of the collection and must contain at least one content file

  • Collection-Raku-Documentation default: 'raku-docs'

• mode is the default mode for the collection, and must be a sub-directory, which must exist and contain a configs sub-directory (note the plural ending). See Mode level configuration below.

  • Collection-Raku-Documentation default: 'Website'

• without-processing as described in Milestones

  • default: False

• no-refresh as described in Milestones

  • default: False

• recompile as described in Milestones

  • default: False

The following are optional keys, together with the defaults

• the allowed extensions for content files. These are provided to the Pod::From::Cache object.

  • default: < rakudoc pod pod6 p6 pm pm6 >

• no-status This option controls whether a progress bar is provided in the terminal

  • default: False

• source-obtain is the array of strings sent to the OS by run to obtain sources, eg git clone and assumes CWD is set to the directory of collection. Without this key, there must already be files in sources.

  • default: ()

• source-refresh is the array of strings run to refresh the sources, assumes CWD set to the directory of sources. No key assumes the sources never change.

  • default: ()

• ignore is a list of files in the sources directory that are not cached. This is a Collection level configuration. The option include-only is Mode level configuration, see below.

  • default: ()

• no-status as described in Milestones

  • default: False

• without-report as described in Milestones

  • default: False

• full-render as described in Milestones

  • default: False

• asset-basename as described in [Asset-cache methods](Asset-cache methods.md)

  • Collection-Raku-Documentation default: 'asset_base'

Mode-level configuration

The mode-level configuration resides in one or more files that are under the configs/ sub-directory of the mode directory. This arrangement is used to allow for configuration to be separated into different named files for ease of management.

The following rules apply:

• If the configs directory for a mode does not exist or is empty, Raku-Doc (collect sub) will fail.

• The Configuration consists of one or more Raku files that each evaluate to a hash.

  • Each Config file in the Configs directory will be evaluated in alphabetical order.

  • Configuration keys may not be over-written. An X::RakuConfig::OverwriteKey exception will be thrown if an attempt is made to redefine a key.

All the following keys are mandatory. Where a key refers to a directory (path), it should normally be relative to the mode sub-directory.

• mode-sources location of the source files for the Collection pages, eg., TOC.

• mode-cache location of the cache files

• the templates subdirectory, which must contain raku files as described in ProcessedPod. These are all passed at the Render milestone directly to the ProcessedPod object.

• destination directory where the output files are rendered

• report-path is the path to which report plugins should output their reports.

• plugin-format defines the format the plugins relate to. Each Mode is specified to produce output in a specific format, and the plugins, which include templates, are related to the format. Published plugins are stored for a particular format.

• plugins-required points to a hash whose keys are milestone names where plugin callables can be applied

  • setup a list of plugin names, see [Plugin management](Plugin management.md), for pre-processing cache contents

  • render plugins used to render Pod::Blocks

  • compilation plugins prepare the ProcessedPod object for collection pages.

  • assets plugins that mark assets created in previous milestones

  • report plugins to test and report on the rendering process

  • completion plugins that define what happens after rendering

  • cleanup plugins if cleanup is needed.

• landing-place is the name of the file that comes first during the completion stage. For example, in a Website, the landing file is usually called index.html

• output-ext is the extension for the output files

• plugin-options is mandatory, see [Plugin level configuration](Plugin level configuration.md) for more information

All optional control flags are False by default. For the Mode configuration they are:

• no-status

• full-render

• without-completion

• without-report

• no-preserved-state

• debug-when

• verbose-when

• no-code-escape

Plugin level configuration

Each plugin has its own configuration (more information in the sections on Plugins). In addition to the mandatory keys, a plugin may have its own configuration data. The configuration data in the plugin directory will be over-written each time a plugin is updated.

In order to provide for preservation of configuration data at the Mode level, the key plugin-options (typically kept in a separate config file) is used. The value of plugin-options is a Hash whose keys are the names of plugins. Each plugin-name key has a value that is a Hash of the keys required by the plugin.

For example, the Collection-Raku-Documentation plugin cro-app has the configuration options :port and :host. The default Collection-Raku-Documentation configuration contains the snippet:

plugin-options => %(
    cro-app => %(
        :port<30000>,
        :host<localhost>,
    ),
),

in a file under the Mode's configs/ directory. These values will over-ride the plugin's default config values.

The plugin should therefore take configuration data from the ProcessedPod instance and not from the config file it is distributed with. This means that if a new plugin is intended to be used in place of an pre-existing one (see [Refresh process](Refresh process.md)), then the developer needs to check the configuration information from the namespace of the replaced name.

The Setup and Completion plugins are passed plugin-options directly because the ProcessedPod instance is out of scope.

plugin-options is a mandatory option in the Mode configuration. It may be set to Empty, viz.,

plugin-options()

in which case, all plugins will use their default options.

The ProcessPod instance is only modified by Render plugins, so if there is plugin configuration data that is needed by another Milestone callable, the plugin should call a blank callable, with empty block and templates.

Control flags

The control flags are also covered in Milestones. Control flags by default are False.

• no-status

No progress status is output at any time.

• without-processing

Setting without-processing to True will skip all the stages except Completion, so long as the destination directories exist.

• no-preserved-state

In order to allow for changes in some source files, or in only mode files, after all the sources have been processed once, the processing state must be archived. This may not be needed in testing or if the archiving takes too long.

Setting no-preserved-state = True prevents storage of state, but also forces without-processing to False, and recompile to True.

• recompile

Controls the updating and caching of the content files. If true, then all files will be recompiled and cached.

A True value is over-ridden by without-processing

Normally False, which allows for only changed files to be processed.

• no-refresh

Prevents the updating of content files, so no changes will be made.

• full-render

Forces all files to be rendered. Even if there are no changes to source files, plugins or templates may be added/changed, thus changing the output, so all files need to be re-rendered.

This flag is set to False if without-processing is True.

• without-report

Normally, report plugin callables report on the final state of the output files. This flag prevents report plugin callables from being loaded or run.

If without-processing is set, then the Report stage is skipped. If, however, the caches do not exist (deleted or first run), then the value of without-processing is ignored and the value of without-report is observed.

• debug-when & verbose-when

ProcessedPod uses debug and verbose, which provide information about which blocks are processed (debug), and the result after the application of the template (verbose). This is a lot of information and generally, it is only one file that is of interest.

These two flags take a string, eg., :debug-when<Introduction.pod6> , and when the filename matches the string, then the debug/verbose flag is set for that file only. (verbose is only effective when debug is True).

• collection-info

Causes collect to produce information about milestones and valid and invalid plugins

• with-only filename

Collect is run only with that filename, which must be in the sources or mode-sources, and is specified like debug-when.

The option takes a string containing the filename. An empty string means all filenames in sources and mode-sources.

• no-code-escape

ProcessedPod has a special flag for turning off escaping in code sections when a highlighter is used to pre-process code. In some cases, the highlighter also does HTML escaping, so RPR has to avoid it.

This has to be done at the Mode level and not left to render plugin callables.

Plugin management

Plugin callables are Raku programs that are executed at specific milestones in the rendering process. The milestone names are given in Milestones above.

The plugins-required key in the Mode's configuration contains a hash with keys whose names are the milestone names. Each key points to a list of plugin names, which are called in the order given.

All plugins must reside within the mode plugins.

All plugin names must be the name of a sub-directory under the plugins subdirectory. Within each plugin sub-directory, there must be a config.raku file containing information for the plugin, and for Collection. If no config.raku files exists, the plugin is not valid and will be skipped.

With the exception of 'render' plugin callables, the config file must contain a key for the milestone type, which points to the program to be called, and when the file is evaluated, it yields a subroutine that takes the parameters needed for the plugin of that milestone. If no key exists with the name of the milestone, then the plugin is not valid.

Plugin's may need other configurable data, which should be kept in the config file for the plugin.

All plugin callables are expected to adhere to no-status and collection-info, which are interpretted as

• no-status if true means 'no output at all', equivalent to a quiet flag

• collection-info if true means 'output extra information' (if need be), equivalent to a verbose flag.

Disabling a plugin

When it's necessary to disable a plugin, this can be done by:

• Removing the plugin name from the plugins-required key of the Mode's config file;

• Renaming / removing the config.raku file name inside the plugin directory

• Renaming / removing the milestone key inside the plugin's config.raku file.

Plugin types

The plugin types are as follows.

Setup

Config hash must contain setup which is the name of a Raku program (a callable) that evaluates to a sub that takes a list of five items, eg.,

sub ( $source-cache, $mode-cache, Bool $full-render, $source-root, $mode-root, %plugin-options, %options ) { ... }

• $source-cache

A Pod::From::Cache+PostCache object containing the pod of the sources files New files can be added to the cache object inside the sub using the .add method, see Sources.

• $mode-cache

Like the above for the mode content files

• $full-render

If True, then the sub should process the cache objects with the .sources method on the cache objects, otherwise with the .list-files method on the cache objects (the .list-files method only provides the files that have changed).

• $source-root

This path must be prepended to any sources added (see below) to the cache, otherwise they will not be added to the destination file.

• $mode-root

Likewise for the mode sources.

• %plugin-options

Has the values of plugin options that over-ride a plugin's own defaults. See [Plugin level configuration](Plugin level configuration.md) for more information.

• %options

Has the values of 'collection-info' and 'no-status' flags.

Render

The Collection plugin-manager calls the ProcessedPod.add-plugin method with the config keys and the path modified to the plugin's subdirectory. The ProcessPod instance is only modified by Render plugins, so if there is plugin configuration data that is needed by another Milestone callable, the plugin should call a blank callable, with empty block and templates.

If the render key is True, no callable is provided, and the plugin name will be added via the .add-plugin method of the ProcessedPod object. See ProcessedPod documentation.

If the render key is a Str, then it is the filename of a Raku callable of the form

sub ( $pr, %options --> Array ) {...}

where

• $pr is a object,

• %options is the same as for Setup, and

• the callable returns a list of triples, with the form (to, from-plug, file)

  • to is the destination under the %config<destination> directory where the asset will be looked for, eg., an image file to be served.

  • from-plug is the name of the plugin in whose directory the asset is contained, where the value myself means the path of the plugin calling the render callable. Actually, 'myself' is the value of Collection::MYSELF.

  • file is the filename local to the source plugin's subdirectory that is to be copied to the destination. This may contain a path relative to the plugin's subdirectory.

  • For example for an HTML file, this would be the relative URL for the src field. Eg., to = 'asset/image', file = 'new-image.png', from-plug = 'myself' and the html would be <img src="asset/image/new-image.png" /> .

Since a render plugin is to be added using the ProcessedPod interface, it must have the custom-raku and template-raku keys defined, even if they evaluate to blank (eg. :custom-raku() ).

So the config file must have:

• render (True | name of callable)

• custom-raku => a Raku program that evaluates to an array of custom blocks (must be set to () if no Raku program )

• template-raku => a Raku program that evaluates to a hash of RakuClosure templates (must be set to () if no Raku program)

It is possible to specify path but it must be relative to the plugin's sub-directory.

Compilation

Note that the structure files are rendered after the compilation stage, BUT the information for rendering the structure files, that is the custom blocks and the templates must accompany a render plugin. Compilation plugin callables are to process the data accumulated during the rendering of the content files, and to make it available for the custom blocks / templates that will be invoked when the structure documents are rendered.

The compilation key must point to a Raku program that delivers a sub object

sub ( $pr, %processed, %options ) { ... }

$pr
is the ProcessedPod object rendering the content files.

%processed
is a hash whose keys are source file names with a hash values containing TOC, Glossary, Links, Metadata, Footnotes, Templates-used structures produced by B.

%options
as for setup

If the return value of the callable is an Array of triplets (as for a Render callable), then assets are transferred from the plugin directory. Any other type of return value is ignored.

Transfer

The transfer key points to a Raku file that evaluates to a

sub ($pr, %processed, %options --> Array ) {...}

%processed
as in Compilation

$pr
as in Compilation

%options
as for Setup

return object
as for the compilation plugin

Report

The report key points to a Raku file that evaluates to a

sub (%processed, @plugins-used, $pr, %options --> Array ) {...}

%processed
as in Compilation

@plugins-used
is an array of Pairs whose key is the milestone and value is a hash of the plugins used and their config parameters.

$pr
as in Compilation

%options
as for Setup

The plugin should return an Array of Pair, where for each Pair .key = (path/)name of the report file with extension, and .value is the text of the report in the appropriate format

The collect sub will write the file to the correct directory.

Completion

The completion key points to a Raku file that evaluates to a

sub ($destination, $landing-place, $output-ext, %plugin-options, %options) {...}

• $destination

is the name of the output path from the mode directory (defined in the mode configuration)

• $landing-place

is the first file to be processed since, eg., for a website, order is not sufficient. name is relative to the destination directory.

• $output-ext

is the output extension.

• %plugin-options

As for Setup

• %options

As for Setup

There is no return value specified for this plugin type.

Plugin updating

The local computer may contain

• More than one collection, eg. a personal website and a Raku documentation collection

• Each Collection may have more than one Mode, eg., a HTML website, an epub producer.

• A collection/mode combination may rely on a previous API of a plugin, which may be broken by a later API.

• A new plugin may have been written as a drop-in replacement for an older version, and the new plugin may have a different name, or auth, or version.

In order to implement this flexibility, the following are specified:

• There is a released plugins directory (see [Released plugins directory](Released plugins directory.md)) to contain all Collection plugins.

• The semantic versioning scheme is mandated for Collection plugins. A version is v<major>.<minor>,<patch>. Changes at the <patch> level do not affect the plugin's functionality. Changes at the <minor> level introduce new functionality, but are backward compatible. Changes at the <major> level break backward compatibility.

• Each distributed plugin is contained in the release directory in a subdirectory named by the plugin name, the auth field, and the major version number (minor and patch numbers are not included because they should not by definition affect the API).

• Each Mode configuration only contains the name of the plugin (without the auth, or version names).

• The developer may define which name/version/auth of a released plugin is to be mapped to the plugin required in the Mode configuration. Thus

  • changes in API can be frozen to earlier versions of the plugin for some concrete Mode.

  • different plugin versions can be used for different collection/modes

  • a differently named plugin can be mapped to a plugin required by a specific collection/mode.

    • Note an alternate plugin, as given by name may also have a non-default auth, so auth may need to be added to plugins.rakuon as well.

• Consequently, all released plugins are defined for

  • a format (eg. html)

  • a major version

  • an auth name

• The mapping information from a released plugin to a Mode is contained in a file in the root of a Collection.

• When the plugins are updated

  • all the latest versions for each Format/Name/Version/Auth are downloaded.

  • a symlink is generated (or if the OS does not allow symlink, the whole directory is copied) from the released version to the directory where each mode expects its plugins to be located.

• Each Collection root directory (the directory containing the topmost config.raku file) will contain the file plugins.rakuon.

• The plugin management tool (PMT)

  • checks if a plugins.rakuon exists. If not, it generates a minimal one.

  • runs through the plugins-required of each Mode in the collection.

  • for each distinct plugin verifies whether

    • the plugin has an entry in plugins.rakuon, in which case

      • the PMT maps (or remaps if the constraint is new) the released plugin name/auth/ver to the plugin-required name using the rules of plugins.rakuon as given in [Mapping released plugins to mode directories](Mapping released plugins to mode directories.md)

    • the plugin does not have an entry, which means it has not yet been mapped, and there are no constraints on the plugin, so the default name/auth/version are used.

Mapping released plugins to mode directories

The file plugins.rakuon contains a hash with the following keys:

• _metadata_. Contains a hash with data for the refresh functionality.

  • collection-plugin-root This contains the name of a directory reachable from the Collection root directory with the released plugins are downloaded.

  • update-behaviour Defines what happens when a new released plugin has an increased Major number. Possible values are:

    • auto All relevant plugin names are updated to the new version, a message is issued

    • conserve Plugins are not updated to the new version, a warning is issued, updating is then plugin by plugin with a -force flag set.

    • Note The update behaviour is not initially implemented.

• Every other toplevel key that meets the plugin naming convenstion is interpreted as a Mode. This means a mode cannot be named _metadata_.

• The Mode key will point to a hash with the keys:

  • _mode_format Each mode may only contain plugins from one Format, eg., html.

  • By the plugin naming rules, a plugin may not be named _mode_format.

  • Every other key in a mode hash must be a plugin name contained in the Mode's plugins-required configuration.

  • There may be zero plugin keys

  • If a plugin in the plugins-required configuration list does not have an entry at this level, then it has not been mapped to a sub-directory of the released-directory.

  • A plugin key that exists must point to a Hash, which must at least contain:

    • mapped => the name of the released plugin

  • The plugin hash may also point to one or more of:

    • name => the name of the alternate released plugin

      • the default name (if the key is absent) is the plugin-required's name

      • if a different name is given, a released plugin is mapped to the required directory in the mode sub-directory

    • major => the major number preceeded by 'v'

      • the default value (if the key is absent) is the greatest released major value

      • A major value is the maximum major value of the full version permitted, thus freezing at that version

    • auth => the auth value

      • the default value (if the key is absent) is 'collection'

  • If there is no distributed plugin for a specified auth | major | name , then an error is thrown.

Some examples:

• The Raku-Collection-Raku-Documentation, Website mode, requires the plugin Camelia. The plugin exists as a HTML format. It has version 1.0.0, and an auth 'collection'. It is distributed as html/camelia_v1_auth_collection. Suppose a version with a new API is created. Then two versions of the plugin will be distributed, including the new one html/camelia_v2_auth_collection.

If the key for camelia in the hash for mode Website only contains an empty version key, then the defaults will be implied and a link (or copy) will be made between the released directory html/camelia_v2_auth_collection and Website/plugins/camelia

• If plugins.rakuon contains the following: Website = %( :FORMA"> {{{ contents }}}

, :camelia( %( :major(1), ) ) > then the link will be between html/camelia_v1_auth_collection and Website/plugins/camelia

• Suppose there is another valid auth string raku-dev and there is a distributed plugin html/camelia_v2_auth_raku-dev, and suppose plugins.rakuon contains the following: Website = %( :FORMA"> {{{ contents }}}

, :camelia( %( :auth, ) ) > then the link will be made between html/camelia_v2_auth_raku-dev and Website/plugins/camelia

• Suppose a different icon is developed called new-camelia by auth raku-dev, then plugins.rakuon may contain the following: Website = %( :FORMA"> {{{ contents }}}

, camelia( %( :name, :auth, ) ) > then a link (copy) is made between html/new-camelia_v2_auth_raku-dev and Website/plugins/camelia

*  Note how the auth must be given for a renaming if there is not a `collection` version of the plugin

Released plugins directory

When Collection is installed, a directory called (by default) $*HOME/.local/share/Collection is created (on a Linux system this will be the same as ~/.local/share/Collection).

If the Environment variable PluginPath is set to a valid path name upon installation, then that will be used instead. But when refresh-collection-plugins is first used, then the non-default name must be supplied.

The directory is initialised to point at the https://github.com/finanalyst/collection-plugins repo and the manifest.rakuon file is downloaded.

If another release directory location is desired, then

• it can be specified when refresh-collection-plugins is first run for each collection, in which case the file plugin.rakuon is given the value for the :collection-plugin-root.

• a custom Raku program for refreshing Collection files can be written, eg., the github repo is to be located in ~/.my_own_collection

use v6.d;
use Collection::RefreshPlugins;
sub MAIN(|c) {
    $Collection::RefreshPlugins::release-dir = "$*HOME/.my_own_collection";
    refresh( |c );
}

Refresh process

The intent is for the released plugins to be held in a single directory (called the released plugins directory), and for the references in a Collection-Mode plugins directory to be links (OS dependent) to the released plugins directory.

The released plugins directory is a Github repository, so doing a git pull will pull in the most recent versions of the plugins. Consequently, each Collection-Mode plugin reference will automatically be updated.

A git pull is therefore one form of a refresh. (TODO if an OS does not have directory links, then this form of refresh will need to be enhanced with a copy operation).

Refresh needs to deal with other situations

• A new plugin name is added to the Collection-Mode's plugins-required list.

  • A link needs to be added to the Collections-Mode's plugins directory.

• A new entry exists for the plugin name in the plugins.rakuon

  • An entry changes the release name associated with the Collection-Mode plugin.

    • If the desired released plugin does not exist, then an Exception is thrown. Default behaviour might lead to an infinite loop.

• A released plugin major version has increased since the last refresh.

  • The default is for a Collection-Mode plugin initially to be linked to the most recent version of the plugin

  • Several behaviours are possible:

    • The default update-behaviour is given in the _metadata_ hash of plugins.rakuon

    • force Leave the existing links in place, issue a warning, update to latest only when forced. Include suggestion to change plugins.rakuon file to suppress warnings.

    • auto (default) Automatically update to the latest version, issuing a message

CLI Plugin Management System

Collection contains the utility collection-refresh-plugins as a (PMS). It is called as follows:

collection-refresh-plugins [-collection=<path>] [-collections='path1 path2 path3']

-collection is the path to a Collection directory. By default it is the Current Working Directory.

-collections is a space delimited list of paths to Collections

When a Collection directory contains a file plugins.rakuon, then the utility will inspect the release directory, updates it, and maps (copy) the most recent plugins according this file. See below for more detail about the specification of plugins.rakuon.

When a Collection does not contain a file <plugins.rakuon>, it generates one from the plugin-required key of each of the config.raku files in each Mode. During this process, the user is prompted for the directory name (relative to the current working directory) of released plugin directory.

Problems and TODO items

Archiving and Minor Changes

In principle, if a small change is made in a source file of a Collection, only the rendered version of that file should be changed, and the Collection pages (eg., the index and the glossaries) updated. The archiving method chosen here is based on Archive::Libarchive and a .7z format. It works in tests where a small quantity of data is stored.

However, when there are many source files (eg., the Raku documentation), the process of restoring state information is significantly longer than re-rendering all the cached files. Consequently, the option no-preserve-state prevents the archiving of processed state. (TODO understanding and optimising the de-archiving process.)

Dump file formatting

The aim is to use PrettyDump instead of <.raku> to transform information into text. However, does not handle Bags properly.

Post-cache methods

Post-cache is a role added to a Pod::From::Load object so that Setup plugin callables can act on Cache'd content by adding pod files to the Cache (perhaps pre-processing primary source files) that will be rendered, masking primary pod files so that they are not rendered, or aliasing primary pod files.

If a secondary source file in the Cache is given a name that is the same as a primary source file, then if the underlying cache object should remain visible, another name (alias) should be given to the file in the Post-cache database.

The Post-cache methods sources, list-files, and pod have the same function and semantics as Pod::From::Cache except that the post-cache database is searched first, and if contents are found there, the contents are returned (which is why post-cache file names hide primary file names). If there is no name in the Post-cache database, then it is passed on to the underlying cache.

multi method add(Str $fn, Array $p --> Pod::From::Cache )

Adds the filename $fn to the cache. $p is expected to be an array of Pod::Blocks, but no check is made. This is intentional to allow the developer flexibility, but then a call to pod( $fn ) will yield an array that is not POD6, which might not be expected.

The invocant is returned, thus allowing add to be chained with mask and alias.

multi method mask(Str $fn --> Pod::From::Cache)

This will add only a filename to the database, and thus mask any existing filename in the underlying cache.

Can be chained.

multi method add-alias(Str $fn, Str :alias! --> Pod::From::Cache)

This will add a filename to the database, with the value of a key in the underlying cache. Chain with mask to prevent the original spelling of the filename in the underlying cache being visible.

Can be chained.

If the alias is already taken, an exception is thrown. This will even occur if the same alias is used for the same cached content file.

method behind-alias(Str $fn --> Str )

Returns the original name of the cached content file, if an alias has been created, otherwise returns the same filename.

method pod(Str $fn)

Will return

• an array of Pod::Block (or other content - beware of adding other content) if the underlying Cache or database have content,

• the array of Pod::Block in an underlying filename, spelt differently

• Nil if there is no content (masking an underlying file in Cache)

• throw a NoPodInCache Exception if there is no pod associated with either the database or the underlying cache. If the original filename is used after an alias have been generated, the Exception will also be thrown.

Asset-cache methods

Asset-cache handles content that is not in Pod6 form. The instance of the Asset-cache class is passed via the plugin-data interface of ProcessedPod, so it is available to all plugin callables after the setup milestone, for example in the plugin callable:

sub render-plugin( $pp ) {
    my $image-manager = $pp.get-data('image-manager');
    ...
    $pp.add-data('custom-block', $image-manager);
}

By creating a name-space in the plugin data section and assigning it the value of the image-manager, the plugin callable can make the image-manager available to templates that get that data, which is a property in parameters called by the name-space.

ProcessedPod provides data from the name-space of a Block, if it exists, as a parameter to the template called for the Block. Note that the default name-space for a block is all lower-case, unless a name-space config option is provided with the Pod Block in the content file.

If a plugin provides an asset (eg., image, jquery script), it needs to provide a render callable that returns the triple so that Collect moves the asset from the plugin directory to the output directory where it can be served. This needs to be done separately if a CSS contains a url for local image.

The basename for the assets is set in the Top level configuration in the option asset-basename

$image-manager is of type Asset-cache, which has the following methods:

    #| the directory base, not included in filenames
    has Str $.basename is rw;
    #| the name of the file being rendered
    Str $.current-file
    #| asset-sources provides a list of all the items in the cache
    method asset-sources
    #| asset-used-list provides a list of all the items that referenced by Content files
    method asset-used-list
    #| asset-add adds an item to the data-base, for example, a transformed image
    method asset-add( $name, $object, :$by = (), :$type = 'image' )
    #| returns name / type / by information in database (not the object blob)
    method asset-db
    #| remove the named asset, and return its metadata
    method asset-delete( $name --> Hash )
    #| returns the type of the asset
    method asset-type( $name --> Str )
    #| if an asset with name and type exists in the database, then it is marked as used by the current file
    #| returns true with success, and false if not.
    method asset-is-used( $asset, $type --> Bool )
    #| brings all assets in directory with given extensions and with type
    #| these are set in the configuration
    multi method asset-slurp( $directory,  @extensions, $type )
    #| this just takes the value of the config key in the top-level configuration
    multi method asset-slurp( %asset-paths )
    #| with type 'all', all the assets are sent to the same output directory
    multi method asset-spurt( $directory, $type = 'all' )
    #| the value of the config key in the mode configuration
    multi method asset-spurt( %asset-paths )

Copyright and License

(c) Copyright, 2021-2022 Richard Hainsworth

LICENSE Artistic-2.0

---

Rendered from README at 2022-12-22T15:48:21Z