Checks, Validations & Autocorrects

Sparge runs checks and autocorrections at multiple pipeline stages. This page is a complete reference of everything it detects and fixes.

At Ingest — autocorrects applied silently

These fixes are applied to every post during ingest. They’re idempotent — running them multiple times produces the same result.

Fix	What it does
`<br/>` normalisation	Converts `<br/>` inside `<pre>` blocks to real newlines
Code block reformatting	Reformats one-liner DRL and XML code blocks with proper newlines
Span-tokenised code	Converts Blogger’s `<span>rule</span><span>"Name"</span>` pattern to `<pre><code>`
`<p><br/>code</p>` blocks	Converts `<p>` blocks containing `<br/>`-separated code to `<pre><code>`
Line-number table code	Converts two-column line-number + code tables to `<pre><code>`
Image link localisation	Downloads `<a href>` image links and replaces with `<img>` tags
Imgur fallback	Geo-blocked imgur CDN addresses are fetched via the Wayback Machine
Wayback fallback	Any image that fails to download is retried via the Wayback Machine

At Enrich — applied on first scan

Enrichment runs automatically the first time you scan a post.

Fix	What it does
YouTube embed replacement	`<iframe>` YouTube embeds replaced with a local thumbnail image + link
Gist inlining	`<script src="gist.github.com/...">` tags fetched and replaced with `<pre><code>`
Code class normalisation	`brush: java`, `lang-java`, etc. normalised to `language-java`
Code block fixes (idempotent)	Same fixes as ingest applied again, in case HTML was modified after ingest

At Scan — HTML issue checks

These are flagged as issues, not auto-fixed. Review each one in the Issues Panel.

Issue type	What it detects	Severity
`no_article`	Post HTML has no `<article>` or `<body>` element — content cannot be located	Error
`data_placeholder`	Lazy-load placeholder image that was never replaced with the real image	Error
`missing_local_image`	Local image file referenced by `<img src>` is missing from the assets directory	Error
`unreplaced_gist`	`<script src="gist.github.com/...">` that wasn’t inlined during enrichment	Error
`empty_embed`	YouTube iframe with no `src` recovered — content is lost	Error
`external_image`	Image still pointing to a remote URL (not localised to assets)	Warning
`tracking_pixel`	1×1 pixel image from a known tracking domain	Warning
`noscript_remnant`	Orphaned `<noscript>` tag containing an image URL	Warning
`wordpress_chrome`	WordPress metadata, admin bars, share buttons, or post navigation	Warning
`missing_image_signal`	Text like “as shown below” with no following image	Warning
`imgur_image`	Image hosted on geo-blocked imgur CDN — replace with Wayback URL	Warning
`layout_spacer_image`	Layout spacer images (e.g. spacer.gif) with no content value — safe to remove	Warning
`md_notation_in_text`	Inline element immediately followed by markdown punctuation — html2text will produce garbled bold/italic	Warning
`suspicious_code_content`	`<pre><code>` block contains HTML-encoded markup — may be a conversion artefact rather than real code	Warning
`code_no_newlines`	Code block using `<br/>` instead of real newlines	Warning
`potential_code_block`	`<p>` block that looks like unformatted code	Warning
`linenumber_table_code`	Two-column line-number + code table not converted during enrichment	Warning

At Generate MD — Markdown validation checks

These run automatically after Markdown generation and when you accept staged edits.

Check	What it detects	Severity
Missing images	Images present in HTML but absent from generated Markdown	Error
Fence breaks	Unclosed or malformed code fences	Error
Garbling	UTF-8 double-encoding artifacts (`ÃÂÃÂ` patterns)	Error
Code block integrity	Fences with no language tag or truncated content	Warning
Cross-validation	Structural cross-check of Markdown against HTML source	Warning

Note: Errors must be resolved before a post can be considered clean. Warnings are advisory — review them but they may be acceptable depending on your content.