SEO Magic

v7.0.16

6 days ago

1. • Your page's real <title> now counts for more in the SEO score, as it is the headline people actually see in search results
2. • The Metadata section of the SEO report no longer grades a page on a <meta name="title"> tag, which no search engine has ever read. Thanks @paulmassen (grav-premium-issues#621)
   • That tag was worth more than half of the Metadata score, so a page with a perfectly good <title> was being marked down to an F simply for leaving it out
   • The tag is still listed in the report for information, with a note explaining that search engines ignore it and that your <title> is what counts

v7.0.15

1 week ago

1. • Fixed a fatal error on the frontend when a page's custom keywords were saved as a plain line of text instead of a list, which happened when the keywords were set in the new admin
   • Custom keywords, custom stopwords, ignored routes, link-checker patterns and the organization's social profiles are all read the same way now, so any of them typed in as a plain comma-separated line works wherever it is used
   • An empty link-checker allow list no longer treats every link as allowed, which had been quietly switching off broken-link reporting
   • The keywords field in the new admin shows the keywords again on pages whose keywords were saved by the classic admin

v7.0.14

1 week ago

1. • The Organization logo in the structured data can be a full URL again. A URL was being handed to the resource locator, which threw on the scheme and took the whole site down with it; a stream such as theme://images/logo.png now resolves to its public URL rather than a filesystem path
   • A site-relative images.default_image such as /user/themes/mytheme/images/og.png is served from the install root; it used to be looked up as a filesystem path and the social image came back 404. The same web path was also handed to is_file() on every page render, which under open_basedir is a warning that Grav turns into an error, so the whole site returned 500

v7.0.13

2 weeks ago

1. • og:type is website for every page that is not a post, where a post is a page with an authored date in its frontmatter. Product, listing and legal pages were all being marked article. A page names its own with seo-magic.opengraph.type, and opengraph.type in the plugin config forces one answer site-wide
   • og:locale is emitted as language_TERRITORY (en_US, not en), as the protocol specifies, with the same normalisation applied to og:locale:alternate. Set it explicitly with opengraph.locale in config or seo-magic.opengraph.locale on a page

v7.0.12

3 weeks ago

1. • Fixed the home page never finding its own crawl data, so it rendered with a bare title and the site-wide description
   • Fixed generated descriptions and keywords staying empty until the cache/seo-magic folder was deleted by hand
   • Fixed a fatal error on 404 pages when a theme calls seomagic_title(), seomagic_description() or seomagic_keywords()

v7.0.11

4 weeks ago

1. • The three scheduled tasks (process, link checker and queue) now run on the schedule you set for them, and appear in the Scheduler under names you can recognise. They were being registered with the schedule in the wrong place, so each one had no schedule of its own, was listed under a meaningless string of letters, and never actually ran

v7.0.10

1 month ago

1. • The width and height given for your social card image are now measured from the image itself, instead of repeating the size set in the plugin options. Those options describe the size the plugin aims for, but only images it resizes itself ever reach it: a webshot screenshot or a default_image is passed straight through at whatever size it arrives at. Sites set to 1200x630 could be telling Facebook, LinkedIn and X one size while serving another, and cards were laid out against numbers the image did not keep. When an image has not been measured yet, the width and height are now left out rather than guessed, and appear once the image has been fetched once.
   • A default_image written as a path on your own site, such as /user/themes/mytheme/images/og.jpg, is now found on disk and measured. Only a full web address was being resolved before, so those sites got no width or height at all.
   • The secure address of the social card image is now sent as og:image:secure_url, the name the Open Graph protocol actually defines. It was previously sent as og:image:secure, which scrapers ignore.

v7.0.9

1 month ago

1. • Added SoftwareApplication structured data for sites whose subject is a piece of software, such as an app, a CMS or a library. Search and answer engines build the candidate set for category queries largely from typed nodes like this, so prose describing the product is not enough on its own. Off by default, and home-page-only when enabled, with per-page overrides so a product landing page can describe itself separately.
2. • The process and link-checker commands no longer refuse to run when you give them the full web address of a sitemap, which already tells them where the site is.
   • When a command does need to be told where the site is, it now suggests running it with --env if this install has environment configurations, instead of only offering to change user/config/system.yaml.

v7.0.8

2 months ago

1. • Fixed saving a page in Admin failing with "Failed to save entry: Encoding JSON failed" on sites whose content uses accented characters, such as Italian, Spanish and Portuguese. The generated summary was being cut to length by byte rather than by character, which could slice an accented character in half and leave text that could not be stored.

v7.0.7

2 months ago

1. • Saving a page in Admin now returns straight away and regenerates its SEO data in the background, instead of making you wait for the page and every link on it to be fetched (grav-premium-issues#610).
   • Links to your own pages are now checked against Grav's own routes rather than fetched one by one, which is dramatically faster and counts several route aliases for the same page only once.
   • The internal and external link concurrency settings now genuinely check links side by side, and are available in the plugin options.
   • Email, phone and in-page anchor links are no longer fetched as if they were web addresses, and the report now shows them as nothing to check.
   • The scheduler can now pick up any page whose SEO data is still pending, and the full site crawl and link check are configurable in the plugin options.
2. • Broken links to your own pages are now reported as broken instead of always coming back as working.
   • The process and link-checker commands, and the scheduled jobs that run them, no longer report "no pages to process" on a site full of pages.
   • Running those commands without a site address configured now says so, instead of quietly fetching every page from http://localhost and storing the results as broken.