Improve performances of the StringSearchService with HTML resources (#162)

`StringSearchService` was sometimes performing very slowly on some HTML resources. The bottleneck was in the parsing from SwiftSoup (third-party dependency). To solve this, we now attempt to parse the resource as strict XML first and fallback on SwiftSoup only when the HTML resource is not a valid XML.

As EPUB resources are supposed to be valid XHTML, this improves performances of EPUB search.

Author: mickael-menu

.github/workflows/build.yml

@@ -13,6 +13,7 @@ jobs:
     env:
       scheme: ${{ 'r2-shared-swift' }}
       platform: ${{ 'iOS Simulator' }}
+      device: ${{ 'iPhone 13' }}
 
     steps:
       - name: Checkout
@@ -30,9 +31,7 @@ jobs:
           rm -rf r2-shared-swift.xcodeproj
       - name: Build
         run: |
-          device=`xcrun xctrace list devices 2>&1 | grep -oE 'iPhone.*?[^\(]+' | head -1 | awk '{$1=$1;print}'`
           xcodebuild build-for-testing -scheme "$scheme" -destination "platform=$platform,name=$device"
       - name: Test
         run: |
-          device=`xcrun xctrace list devices 2>&1 | grep -oE 'iPhone.*?[^\(]+' | head -1 | awk '{$1=$1;print}'`
           xcodebuild test-without-building -scheme "$scheme" -destination "platform=$platform,name=$device"

CHANGELOG.md

@@ -10,6 +10,11 @@ All notable changes to this project will be documented in this file.
 
 * Support for Paragraph Margins user setting.
 
+### Fixed
+
+* Improved performances of the search service used with EPUB.
+
+
 ## [2.1.0]
 
 ### Added

r2-shared-swift/Fetcher/Resource/ResourceContentExtractor.swift

@@ -54,19 +54,47 @@ public class _DefaultResourceContentExtractorFactory: _ResourceContentExtractorF
 /// notice. Use with caution.
 class _HTMLResourceContentExtractor: _ResourceContentExtractor {
 
+    private let xmlFactory = DefaultXMLDocumentFactory()
+
     func extractText(of resource: Resource) -> ResourceResult<String> {
         resource.readAsString()
-            .flatMap { html in
+            .flatMap { content in
                 do {
-                    var text = try SwiftSoup.parse(html).body()?.text() ?? ""
+                    // First try to parse a valid XML document, then fallback on SwiftSoup, which is slower.
+                    var text = parse(xml: content)
+                        ?? parse(html: content)
+                        ?? ""
+                    
                     // Transform HTML entities into their actual characters.
                     text = try Entities.unescape(text)
+                    
                     return .success(text)
 
                 } catch {
                     return .failure(.wrap(error))
                 }
             }
     }
-
-}
+    
+    // Parse the HTML resource as a strict XML document.
+    //
+    // This is much more efficient than using SwiftSoup, but will fail when encountering
+    // invalid HTML documents.
+    private func parse(xml: String) -> String? {
+        guard let document = try? xmlFactory.open(string: xml, namespaces: [
+                XMLNamespace(prefix: "xhtml", uri: "http://www.w3.org/1999/xhtml")
+            ])
+        else {
+            return nil
+        }
+        
+        return document.first("/xhtml:html/xhtml:body")?.textContent
+    }
+    
+    // Parse the HTML resource with SwiftSoup.
+    //
+    // This may be slow but will recover from broken HTML documents.
+    private func parse(html: String) -> String? {
+        return try? SwiftSoup.parse(html).body()?.text()
+    }
+}

r2-shared-swift/Toolkit/XML/Fuzi.swift

@@ -18,6 +18,7 @@ final class FuziXMLDocument: XMLDocument, Loggable {
 
     init(string: String, namespaces: [XMLNamespace]) throws {
         self.document = try Fuzi.XMLDocument(string: string)
+        document.definePrefixes(namespaces)
     }
 
     lazy var documentElement: XMLElement? =