trek-rs 0.2.1

//! Constants used throughout Trek

use once_cell::sync::Lazy;
use regex::Regex;

/// Mobile viewport width for style evaluation
pub const MOBILE_WIDTH: u32 = 600;

/// Entry point selectors for finding main content
pub const ENTRY_POINT_ELEMENTS: &[&str] = &[
    "#post",
    ".post-content",
    ".article-content",
    "#article-content",
    ".article_post",
    ".article-wrapper",
    ".entry-content",
    ".content-article",
    ".post",
    ".markdown-body",
    "article",
    "[role=\"article\"]",
    "main",
    "[role=\"main\"]",
    "body", // ensures there is always a match
];

/// Block element tag names
pub const BLOCK_ELEMENTS: &[&str] = &[
    "div", "section", "article", "main", "aside", "header", "footer", "nav", "content",
];

/// Elements that should be preserved during flattening
pub const PRESERVE_ELEMENTS: &[&str] = &[
    "pre",
    "code",
    "table",
    "thead",
    "tbody",
    "tr",
    "td",
    "th",
    "ul",
    "ol",
    "li",
    "dl",
    "dt",
    "dd",
    "figure",
    "figcaption",
    "picture",
    "details",
    "summary",
    "blockquote",
    "form",
    "fieldset",
];

/// Inline elements that should not be unwrapped
pub const INLINE_ELEMENTS: &[&str] = &[
    "a",
    "span",
    "strong",
    "em",
    "i",
    "b",
    "u",
    "code",
    "br",
    "small",
    "sub",
    "sup",
    "mark",
    "date",
    "del",
    "ins",
    "q",
    "abbr",
    "cite",
    "relative-time",
    "time",
    "font",
];

/// Exact selectors to remove
pub const EXACT_SELECTORS: &[&str] = &[
    // scripts, styles
    "noscript",
    "script:not([type^=\"math/\"])",
    "style",
    "meta",
    "link",
    // ads
    ".ad:not([class*=\"gradient\"])",
    "[class^=\"ad-\" i]",
    "[class$=\"-ad\" i]",
    "[id^=\"ad-\" i]",
    "[id$=\"-ad\" i]",
    "[role=\"banner\" i]",
    "[alt*=\"advert\" i]",
    ".promo",
    ".Promo",
    "#barrier-page",
    ".alert",
    // comments
    "[id=\"comments\" i]",
    "[id=\"comment\" i]",
    // header, nav
    "header",
    ".header:not(.banner)",
    "#header",
    "#Header",
    "#banner",
    "#Banner",
    "nav",
    ".navigation",
    "#navigation",
    ".hero",
    "[role=\"navigation\" i]",
    "[role=\"dialog\" i]",
    "[role*=\"complementary\" i]",
    "[class*=\"pagination\" i]",
    ".menu",
    "#menu",
    "#siteSub",
    ".previous",
    // footer
    "footer",
    // inputs, forms, elements
    ".aside",
    "aside",
    "button",
    "canvas",
    "date",
    "dialog",
    "fieldset",
    "form",
    "input:not([type=\"checkbox\"])",
    "label",
    "option",
    "select",
    "textarea",
    "time",
    "relative-time",
    // hidden
    "[hidden]",
    "[aria-hidden=\"true\"]:not([class*=\"math\"])",
    "[style*=\"display: none\"]:not([class*=\"math\"])",
    "[style*=\"display:none\"]:not([class*=\"math\"])",
    "[style*=\"visibility: hidden\"]",
    "[style*=\"visibility:hidden\"]",
    ".hidden",
    ".invisible",
    // iframes
    "instaread-player",
    "iframe:not([src*=\"youtube\"]):not([src*=\"youtu.be\"]):not([src*=\"vimeo\"]):not([src*=\"twitter\"]):not([src*=\"x.com\"]):not([src*=\"datawrapper\"])",
    // newsletter
    "#newsletter",
    "#Newsletter",
    ".subscribe",
    // sidebar
    ".sidebar",
    ".Sidebar",
    "#sidebar",
    "#Sidebar",
    "#sitesub",
];

/// Attributes to test for partial selectors
pub const TEST_ATTRIBUTES: &[&str] = &[
    "class",
    "id",
    "data-test",
    "data-testid",
    "data-test-id",
    "data-qa",
    "data-cy",
];

/// Partial selector patterns for removal
pub const PARTIAL_SELECTORS: &[&str] = &[
    "a-statement",
    "access-wall",
    "activitypub",
    "actioncall",
    "addcomment",
    "advert",
    "adlayout",
    "ad-tldr",
    "ad-placement",
    "ads-container",
    "_ad_",
    "after_content",
    "after_main_article",
    "afterpost",
    "alert-box",
    "appendix",
    "_archive",
    "around-the-web",
    "article-author",
    "article-badges",
    "article-banner",
    "article-bottom",
    "article-category",
    "article-card",
    "article-citation",
    "article_date",
    "article-date",
    "article-end ",
    "article_header",
    "article-header",
    "article__header",
    "article__info",
    "article-info",
    "article-meta",
    "article_meta",
    "article__meta",
    "article-tags",
    "article_tags",
    "articletitle",
    "article-title",
    "article_title",
    "author-bio",
    "author-box",
    "author-info",
    "author_info",
    "back-to-top",
    "banner",
    "bio-block",
    "blog-pager",
    "bookmark-",
    "-bookmark",
    "bottomnav",
    "bottom-of-article",
    "breadcrumb",
    "btn-",
    "-btn",
    "byline",
    "comments",
    "comment-box",
    "comment-form",
    "contact-",
    "content-card",
    "copyright",
    "dateline",
    "date-header",
    "disclaimer",
    "disqus",
    "donate",
    "dropdown",
    "email-signup",
    "entry-date",
    "entry-meta",
    "entry-title",
    "facebook",
    "feedback",
    "footer",
    "further-reading",
    "header-logo",
    "hidden-print",
    "instagram",
    "jump-to-",
    "keep-reading",
    "latest-content",
    "license-",
    "links-grid",
    "loading",
    "login",
    "masthead",
    "menu-",
    "metadata",
    "more-articles",
    "more-like-this",
    "most-read",
    "nav-",
    "navigation",
    "newsletter",
    "next-",
    "popular",
    "post-author",
    "post-date",
    "post-meta",
    "post-tags",
    "post-title",
    "prev-",
    "print-",
    "privacy",
    "profile",
    "promo",
    "read-more",
    "read-next",
    "recent-",
    "recommend",
    "register",
    "related",
    "rss-",
    "search-",
    "share-",
    "sharing",
    "sidebar",
    "sign-in",
    "sign-up",
    "signup",
    "skip-",
    "social",
    "sponsor",
    "subscribe",
    "tag-",
    "tags",
    "terms",
    "timestamp",
    "trending",
    "twitter",
    "widget",
];

/// Footnote inline reference selectors
pub const FOOTNOTE_INLINE_REFERENCES: &[&str] = &[
    "sup.reference",
    "cite.ltx_cite",
    "sup[id^=\"fnr\"]",
    "span[id^=\"fnr\"]",
    "span[class*=\"footnote_ref\"]",
    "span.footnote-link",
    "a.citation",
    "a[id^=\"ref-link\"]",
    "a[href^=\"#fn\"]",
    "a[href^=\"#cite\"]",
    "a[href^=\"#reference\"]",
    "a[href^=\"#footnote\"]",
    "a[href^=\"#r\"]",
    "a[href^=\"#b\"]",
    "a[href*=\"cite_note\"]",
    "a[href*=\"cite_ref\"]",
    "a.footnote-anchor",
    "span.footnote-hovercard-target a",
    "a[role=\"doc-biblioref\"]",
    "a[id^=\"fnref\"]",
    "a[id^=\"ref-link\"]",
];

/// Footnote list selectors
pub const FOOTNOTE_LIST_SELECTORS: &[&str] = &[
    "div.footnote ol",
    "div.footnotes ol",
    "div[role=\"doc-endnotes\"]",
    "div[role=\"doc-footnotes\"]",
    "ol.footnotes-list",
    "ol.footnotes",
    "ol.references",
    "ol[class*=\"article-references\"]",
    "section.footnotes ol",
    "section[role=\"doc-endnotes\"]",
    "section[role=\"doc-footnotes\"]",
    "section[role=\"doc-bibliography\"]",
    "ul.footnotes-list",
    "ul.ltx_biblist",
    "div.footnote[data-component-name=\"FootnoteToDOM\"]",
];

/// Elements allowed to be empty
pub const ALLOWED_EMPTY_ELEMENTS: &[&str] = &[
    "area", "audio", "base", "br", "circle", "col", "defs", "ellipse", "embed", "figure", "g",
    "hr", "iframe", "img", "input", "line", "link", "mask", "meta", "object", "param", "path",
    "pattern", "picture", "polygon", "polyline", "rect", "source", "stop", "svg", "td", "th",
    "track", "use", "video", "wbr",
];

/// Attributes to preserve in output
pub const ALLOWED_ATTRIBUTES: &[&str] = &[
    "alt",
    "allow",
    "allowfullscreen",
    "aria-label",
    "checked",
    "colspan",
    "controls",
    "data-latex",
    "data-src",
    "data-srcset",
    "data-lang",
    "dir",
    "display",
    "frameborder",
    "headers",
    "height",
    "href",
    "lang",
    "role",
    "rowspan",
    "src",
    "srcset",
    "title",
    "type",
    "width",
    // MathML attributes
    "accent",
    "accentunder",
    "align",
    "columnalign",
    "columnlines",
    "columnspacing",
    "columnspan",
    "data-mjx-texclass",
    "depth",
    "displaystyle",
    "fence",
    "frame",
    "framespacing",
    "linethickness",
    "lspace",
    "mathsize",
    "mathvariant",
    "maxsize",
    "minsize",
    "movablelimits",
    "notation",
    "rowalign",
    "rowlines",
    "rowspacing",
    "rowspan",
    "rspace",
    "scriptlevel",
    "separator",
    "stretchy",
    "symmetric",
    "voffset",
    "xmlns",
];

/// Additional attributes allowed in debug mode
pub const ALLOWED_ATTRIBUTES_DEBUG: &[&str] = &["class", "id"];

/// Lazy regex for combined partial selectors
pub static PARTIAL_SELECTORS_REGEX: Lazy<Regex> = Lazy::new(|| {
    let pattern = PARTIAL_SELECTORS.join("|");
    Regex::new(&pattern).expect("Failed to compile partial selectors regex")
});

/// Content indicator patterns
pub const CONTENT_INDICATORS: &[&str] = &[
    "admonition",
    "article",
    "content",
    "entry",
    "image",
    "img",
    "font",
    "figure",
    "figcaption",
    "pre",
    "main",
    "post",
    "story",
    "table",
];

/// Navigation indicator patterns
pub const NAVIGATION_INDICATORS: &[&str] = &[
    "advertisement",
    "all rights reserved",
    "banner",
    "cookie",
    "comments",
    "copyright",
    "follow me",
    "follow us",
    "footer",
    "header",
    "homepage",
    "login",
    "menu",
    "more articles",
    "more like this",
    "most read",
    "nav",
    "navigation",
    "newsletter",
    "popular",
    "privacy",
    "recommended",
    "register",
    "related",
    "responses",
    "share",
    "sidebar",
    "sign in",
    "sign up",
    "signup",
    "social",
    "sponsored",
    "subscribe",
    "terms",
    "trending",
];

/// Non-content class/id patterns
pub const NON_CONTENT_PATTERNS: &[&str] = &[
    "ad",
    "banner",
    "cookie",
    "copyright",
    "footer",
    "header",
    "homepage",
    "menu",
    "nav",
    "newsletter",
    "popular",
    "privacy",
    "recommended",
    "related",
    "rights",
    "share",
    "sidebar",
    "social",
    "sponsored",
    "subscribe",
    "terms",
    "trending",
    "widget",
];