@sdxc/spam 2026.10.6 → 2026.10.10

package/dist/lib/text.js

@@ -11,7 +11,7 @@
  */
 const URL_PATTERN = /\b(?:https?:\/\/(?:\[[0-9a-f:.]+\])?|www\.)[^\s"'<>()[\]]*/gi;
 /** Punctuation a sentence puts after a URL, which is never part of it. */
-const TRAILING_PUNCTUATION = /[.,;:!?]+$/;
+const TRAILING_PUNCTUATION = new Set([".", ",", ";", ":", "!", "?"]);
 /**
  * The distinct links in `text`, parsed. A `www.` link is read as `http://`; a match that fails to
  * parse is dropped, so every entry has a host.
@@ -22,7 +22,7 @@ export function extractLinks(text) {
     let seen = new Set();
     let links = [];
     for (let match of text.matchAll(URL_PATTERN)) {
-        let raw = match[0].replace(TRAILING_PUNCTUATION, "");
+        let raw = trimTrailingPunctuation(match[0]);
         let href = /^www\./i.test(raw) ? `http://${raw}` : raw;
         if (!URL.canParse(href))
             continue;
@@ -34,6 +34,16 @@ export function extractLinks(text) {
     }
     return links;
 }
+/**
+ * `raw` without the sentence punctuation that ends it. Walking back from the end keeps the cost
+ * linear in the match however long a run of punctuation a submission packs into it.
+ */
+function trimTrailingPunctuation(raw) {
+    let end = raw.length;
+    while (end > 0 && TRAILING_PUNCTUATION.has(raw.charAt(end - 1)))
+        end -= 1;
+    return raw.slice(0, end);
+}
 /** The words of `text` with its links removed, split on anything that is not a letter or digit. */
 export function extractWords(text) {
     return text

package/dist/rules/link-syntax.js

@@ -1,9 +1,68 @@
-/** A BBCode link, `[url=…]` or `[url]…[/url]`. */
-const BBCODE_LINK = /\[url(?:=[^\]]*)?\]/i;
-/** An HTML anchor with an `href`. */
-const HTML_LINK = /<a\s[^>]*href\s*=/i;
-/** A Markdown inline link to an absolute URL. */
-const MARKDOWN_LINK = /\[[^\]]+\]\(\s*https?:\/\//i;
+/** The opening of a BBCode link, `[url]` or `[url=`; the latter needs a `]` somewhere after it. */
+const BBCODE_OPEN = /\[url[=\]]/i;
+/** The opening of an HTML anchor, tried only where a `<` stands. */
+const ANCHOR_OPEN = /<a\s/iy;
+/** An `href` attribute, tried only inside an open anchor tag. */
+const HREF_ATTRIBUTE = /href\s*=/iy;
+/** The absolute-URL target that turns a bracketed label into a Markdown link. */
+const MARKDOWN_TARGET = /\(\s*https?:\/\//iy;
+/**
+ * Whether `content` holds a BBCode link: `[url]`, or `[url=` with a `]` after it. Only the first
+ * opening is checked, since a `]` after any later one also follows the first.
+ */
+function hasBbcodeLink(content) {
+    let open = BBCODE_OPEN.exec(content);
+    if (open === null)
+        return false;
+    return open[0].endsWith("]") || content.includes("]", open.index + open[0].length);
+}
+/**
+ * Whether an HTML anchor tag in `content` carries an `href`. One pass tracks whether the scan is
+ * inside an anchor tag opened since the last `>`, so the cost stays linear in the submission.
+ */
+function hasHtmlLink(content) {
+    let inAnchor = false;
+    for (let at = 0; at < content.length; at += 1) {
+        let character = content.charAt(at);
+        if (character === ">") {
+            inAnchor = false;
+        }
+        else if (character === "<") {
+            ANCHOR_OPEN.lastIndex = at;
+            if (ANCHOR_OPEN.test(content))
+                inAnchor = true;
+        }
+        else if (inAnchor && (character === "h" || character === "H")) {
+            HREF_ATTRIBUTE.lastIndex = at;
+            if (HREF_ATTRIBUTE.test(content))
+                return true;
+        }
+    }
+    return false;
+}
+/**
+ * Whether `content` holds a Markdown inline link to an absolute URL: a non-empty `[label]` followed
+ * by `(http…`. One pass remembers the first `[` since the last `]`, so the cost stays linear.
+ */
+function hasMarkdownLink(content) {
+    let open = -1;
+    for (let at = 0; at < content.length; at += 1) {
+        let character = content.charAt(at);
+        if (character === "[") {
+            if (open === -1)
+                open = at;
+        }
+        else if (character === "]") {
+            if (open !== -1 && at - open > 1) {
+                MARKDOWN_TARGET.lastIndex = at + 1;
+                if (MARKDOWN_TARGET.test(content))
+                    return true;
+            }
+            open = -1;
+        }
+    }
+    return false;
+}
 /**
  * Scores BBCode links in any format, HTML anchors outside `html`, and Markdown links in `text`.
  * Each kind scores once however many links use it.
@@ -21,21 +80,21 @@ export function linkSyntax(options = {}) {
             let format = submission.format ?? "text";
             let content = submission.content;
             let signals = [];
-            if (BBCODE_LINK.test(content)) {
+            if (hasBbcodeLink(content)) {
                 signals.push({
                     check: "link-syntax.bbcode",
                     score: bbcodeScore,
                     detail: "BBCode link markup",
                 });
             }
-            if (format !== "html" && HTML_LINK.test(content)) {
+            if (format !== "html" && hasHtmlLink(content)) {
                 signals.push({
                     check: "link-syntax.html",
                     score: htmlScore,
                     detail: `HTML anchors in a ${format} field`,
                 });
             }
-            if (format === "text" && MARKDOWN_LINK.test(content)) {
+            if (format === "text" && hasMarkdownLink(content)) {
                 signals.push({
                     check: "link-syntax.markdown",
                     score: markdownScore,

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@sdxc/spam",
-	"version": "2026.10.6",
+	"version": "2026.10.10",
 	"description": "Spam scoring for user content: local rules, reputation checks and a classifier",
 	"license": "MIT",
 	"type": "module",
@@ -15,12 +15,12 @@
 		"./workers-ai": "./dist/workers-ai.js"
 	},
 	"dependencies": {
-		"@sdxc/doh": "2026.10.6",
-		"@sdxc/email-address": "2026.10.6",
+		"@sdxc/doh": "2026.10.10",
+		"@sdxc/email-address": "2026.10.10",
 		"@sdxc/result": "2026.10.6",
 		"remix": "3.0.0"
 	},
-	"gitHead": "b1067eb20fc47ec51ff8c4de47d07dc3f9c00466",
+	"gitHead": "b61f19c59866708a2d89a5a8f4655ebbc16941b0",
 	"publishConfig": {
 		"access": "public"
 	},