Word counts and reading time

Every declared Markdown input exposes a deterministic Unicode word count. Templates can use the fact without deciding which content file is a page's primary body:

{{ page.markdown.main.word_count }}

page.content.main remains the value rendered through markdown; the parallel page.markdown.main view contains facts derived from that original source.

Enable an estimate

Reading time is optional because it depends on a site's audience and language. Choose one positive words-per-minute rate for each configured locale:

[reading_time.words_per_minute]
en = 236
de = 260

The value then becomes a whole-minute estimate, rounded up:

reading-time = about { $minutes ->
   [one] one minute
  *[other] { $minutes } minutes
} read

{% if page.markdown.main.reading_time %}
  <p>{{ t(id="reading-time", minutes=page.markdown.main.reading_time) }}</p>
{% endif %}

Keep the label in Fluent. A template may present the count, the estimate, both, or neither. Without [reading_time], reading_time is null; word_count remains available. A zero-word input has an estimate of zero minutes when the policy is enabled.

Choosing a rate

A reading-time value is an estimate, not a claim about a reader. Text difficulty, familiarity, age, and whether someone is studying rather than reading all affect it. The following non-normative starting points are mean silent-reading rates reported by a review of the available studies. They apply to the studies' word units and populations, not to every site or reader.

The review and meta-analysis by Marc Brysbaert is the source for these values. It reports substantial variation, including an English non-fiction range of roughly 175 to 300 words per minute. Start with a relevant value, then adjust it when you know more about the site's readers.

Heine does not supply these values automatically. A multilingual site must choose every configured locale's rate, so an English convention never becomes an implicit policy for another language.

What Heine counts

The count comes from parsed Markdown before templates render it. It includes prose in headings, paragraphs, quotations, lists, tables, links, image alternatives, and footnote definitions. It excludes Markdown syntax, URLs, comments, recognized math expressions, and fenced code blocks. Code is intentionally excluded because comprehension time, rather than reading speed, normally determines how long it takes to understand.

Trusted raw HTML contributes its textual content, but not markup, attributes, URLs, comments, or script, style, and template contents. A color-scheme picture contributes its authored alternative text and caption. Each segmented Markdown input is counted once as its original source, independent of how a template renders its segments.

Heine uses Unicode default word segmentation. This is deterministic across hosts, but it is not a language-specific lexical tokenizer. Scripts such as Chinese, Japanese, Thai, Lao, and Khmer may need tailored segmentation for a linguistic word count. Choose rates for the unit Heine actually reports rather than copying a benchmark based on a different tokenizer.

See the reference for configuration and template field constraints.