Client side in browser rendering

The idea is to generate texts client side, in a browser, without node.js.

A use case is for business intelligence application, where charts are generated on the fly dynamically in the browser, and with RosaeNLG one could do the same for texts.

Limits

Generating texts in the browser works perfectly. There are 2 limitations to be aware of:

The default bundled packages per language do not contain most of their linguistic resources to make them lighter. The linguistic resources have to be bundled inside it at compile time: the ones which are effectively used in templates are collected automatically during compilation, and you can also indicate expressly the other ones.
The simplified syntax <…> is desactivated for fr_FR, de_DE and it_IT: you cannot use it in your templates when doing client side rendering in those languages. The reason is that this feature requires lefff-helper module (for French), german-dict-helper (for German), and morph-it-helper for Italian and that they require resource files that would be too large to run in a browser.

How to make it work

A possible pipeline is the following:

compile your templates using compileFileClient (put embedResources to true)
package them using browserify
import the compiled template and the proper rosaenlg_*_*.js lib in your webpage
generate texts in your webpage just calling the template

You should use rosaenlg-packager to ease the two first steps.

In your html file, import the proper packages. rosaenlg_tiny_en_US_VERSION.js is the compiled and packaged template.

<script src="compiled_en_US.js"></script>
<script src="rosaenlg_tiny_en_US_0.18.9.js"></script>

You can also use a CDN: https://unpkg.com/rosaenlg@VERSION/dist/rollup/rosaenlg_tiny_LANG_VERSION_comp.js, for instance https://unpkg.com/rosaenlg@1.20.2/dist/rollup/rosaenlg_tiny_en_US_1.20.2_comp.js

In your html file, run this JavaScript to generate texts:

let rendered = template({
  util: new rosaenlg_en_US.NlgLib({language: 'en_US'}),
  data: ...
});

template is the name of the function rendering the compiled template; it an option when calling rosaenlg-packager.

For a complete technical POC that effectively generates texts in a browser using RosaeNLG, see rosaenlg-browser-poc module.

Bundling linguistic resources

Linguistic resources (extensive lists of verbs, words, adjectives) are not bundled directly inside the packaged versions for browsers: if they were, the packages would be too large to be transfered to a browser. Moreover, a project typically only uses effectively a few verbs/words/adjectives among the whole available list.

To bundle linguistic resources:

When compiling, you must provide the language parameter. While you generally don’t have to indicate language at compile time, here it is necessary so that RosaeNLG knows how to resolve resources to bundle them.
When compiling, you must set embedResources to true so that RosaeNLG knows it must embed resources in the compiled template.
Some/most of the verbs/words/adjectives are properly detected at compile time and properly embedded automatically. For instance in #[+verb(…, {'verb' : 'danser', tense:'FUTUR'} )], danser will be detected.
For the verbs/words/adjectives which are not properly detected (for instance in #[+verb(…, getVerbXXX() )] - and you will see it at runtime when rendering fails), you can explicitly indicate the resources to embed. Use verbs, words, adjectives parameters in the compile options; all of them are lists

For undetected resources, you can as well add a dummy mixin where resources will be properly detected:

mixin forResourcesOnly
  //- will add all words and verbs and adjectives (but don't call the génie!)
  | #[+subjectVerbAdj(['lampe', 'génie'], ['sembler', 'voir'], ['somptueux', 'beau'])]

Table 1. Linguistic resources which can be bundled per language
Language	Linguistic resources
`en_US`	verbs, words
`fr_FR`	verbs, words, adjectives
`de_DE`	verbs, words, adjectives
`it_IT`	verbs, words, adjectives
`es_ES`	verbs, words, adjectives

A complete example in French:

const templateVerb = `
p
  | il #[+verb(getAnonMS(), {verb: 'chanter', tense:'FUTUR'} )]
`;

const compiled = rosaenlgPug.compileClient(templateVerb, {
  language: 'fr_FR',
  compileDebug: false,
  embedResources: true
});

const compiledFct = new Function('params', `${compiled}; return template(params);`);

let rendered = compiledFct({
  util: new NlgLib({language: 'fr_FR'})
});

will output il chante, without requiring any linguistic resource at runtime.

When a resource has been forgotten, execution will fail; watch for logs as they should indicate the exact name of the forgotten resource (e.g. could not read xxxx on disk: …).

Auxiliary verbs (être and avoir for French, sein, haben and werden for German) are always automatically included.