anchor-checker.php: a simple script to check HTML self-link validity

[Gwern Branwen <gwern-v26ZT+9V8bxeoWH0uzbU5w@public.gmane.org>]

When writing long complicated Pandoc documents over years, it is easy
for section or anchor self-links to drift and become invalid.
Pandoc/hakyll will not warn you about links being broken, they do not
break in any visible ways or affect validation, and it's surprisingly
difficult to get any linkchecking tool to warn you about them
specifically. (The W3C linkchecking tool has no option to check only
anchors and insists on checking all external links, while
'linkchecker' has a plugin for this I am told is broken and also no
way to avoid checking all external links - this renders both tools
completely infeasible for regularly checking a website the size of
gwern.net, where a linkchecker run may take several days to finish.)
This is despite being really quite a simple task: get all hrefs
starting with '#', check that an ID corresponds, print out any ones
missing.

dbohdan wrote a PHP script for me to check gwern.net pages, which has
worked well and exposed at least 20 erroneous anchors I've fixed, and
which is fast enough to include in the site sync script so new errors
will be picked up immediately. This may be useful to other
Pandoc/hakyll users.

Source: https://github.com/gwern/gwern.net/blob/master/build/anchor-checker.php

    #! /usr/bin/env php
    <?php
    // Check anchors in HTML files.  Only checks anchors local to each document.
    // Anchors prefixed with a filename are ignored even if they refer to the
    // same file.  Anchors with no element with the corresponding fragment ID
    // are written to stderr prefixed with the filename.
    //
    // Usage: anchor-checker.php [FILE]...
    //
    // To the extent possible under law, D. Bohdan has waived all copyright and
    // related or neighboring rights to this work.
    //
    // Date: 2021-05-24.
    // Requirements: PHP 7.x with the standard DOM module.

    error_reporting(E_ALL);

    function main($files) {
        $exit_code = 0;

        foreach ($files as $file) {
            $bad_anchors = check_file($file);
            foreach ($bad_anchors as $a) {
                fprintf(STDERR, "%s\t%s\n", $file, $a);
                $exit_code = 1;
            }
        }

        exit($exit_code);
    }

    function check_file($file) {
        $html = file_get_contents($file);
        // An ugly hack to get around missing HTML5 support tripping
up the parser.
        $html = preg_replace("/<wbr>/", "", $html);

        if (preg_match("/^\s*$/", $html)) return [];

        $dom = new DOMDocument();
        $dom->loadHTML($html, LIBXML_NOERROR | LIBXML_NOWARNING);

        return check_document($dom);
    }

    function check_document($dom) {
        $ids = (new DOMXpath($dom))->query("//@id");
        $id_set = array();

        foreach ($ids as $id) {
            $id_set["#" . $id->value] = true;
        }

        $bad_anchors = array();
        $hrefs = (new DOMXpath($dom))->query("//a/@href");
        foreach ($hrefs as $href) {
            $value = trim($href->value);

            if (substr($value, 0, 1) !== "#") continue;

            if (!array_key_exists($value, $id_set)) {
                $bad_anchors[] = $value;
            }
        }

        return $bad_anchors;
    }

    main(array_slice($argv, 1));

This can be used as a post-compilation check like
`static/build/anchor-checker.php ./_site/"$HTML"` or what have you.

-- 
gwern
https://www.gwern.net