From b4b5d802c9469fabaf469f059f1f6f6480a949bc Mon Sep 17 00:00:00 2001
From: Jan Lukas Gernert <jangernert@gmail.com>
Date: Fri, 21 Apr 2023 08:46:10 +0200
Subject: [PATCH] only serialize root node

---
 article_scraper/src/article.rs                | 13 +++++--
 article_scraper/src/clean.rs                  | 34 +++++++------------
 article_scraper/src/full_text_parser/mod.rs   | 28 ++++++++++-----
 .../src/full_text_parser/readability/mod.rs   |  2 ++
 .../src/full_text_parser/readability/tests.rs |  2 ++
 5 files changed, 47 insertions(+), 32 deletions(-)

diff --git a/article_scraper/src/article.rs b/article_scraper/src/article.rs
index 3408e15..994265f 100644
--- a/article_scraper/src/article.rs
+++ b/article_scraper/src/article.rs
@@ -1,5 +1,5 @@
 use chrono::{DateTime, Utc};
-use libxml::tree::{Document, SaveOptions};
+use libxml::tree::{Document, SaveOptions, Node};
 use std::fs::File;
 use std::io::{Error, ErrorKind, Write};
 use std::path::PathBuf;
@@ -12,10 +12,11 @@ pub struct Article {
     pub date: Option<DateTime<Utc>>,
     pub thumbnail_url: Option<String>,
     pub document: Option<Document>,
+    pub root_node: Option<Node>,
 }
 
 impl Article {
-    pub fn get_content(&self) -> Option<String> {
+    pub fn get_doc_content(&self) -> Option<String> {
         // serialize content
         let options = SaveOptions {
             format: true,
@@ -32,6 +33,14 @@ impl Article {
             .map(|doc| doc.to_string_with_options(options))
     }
 
+    pub fn get_content(&self) -> Option<String> {
+        if let (Some(document), Some(root)) = (self.document.as_ref(), self.root_node.as_ref()) {
+            Some(document.node_to_string(root))
+        } else {
+            None
+        }
+    }
+
     pub fn save_html(&self, path: &PathBuf) -> Result<(), Error> {
         if let Some(ref html) = self.get_content() {
             if let Ok(()) = std::fs::create_dir_all(path) {
diff --git a/article_scraper/src/clean.rs b/article_scraper/src/clean.rs
index 37f4bf2..0b8519c 100644
--- a/article_scraper/src/clean.rs
+++ b/article_scraper/src/clean.rs
@@ -1,7 +1,8 @@
-use libxml::tree::SaveOptions;
+use libxml::tree::Node;
 use reqwest::Url;
 
 use crate::full_text_parser::error::FullTextParserError;
+use crate::util::Util;
 use crate::{FtrConfigEntry, FullTextParser};
 
 pub fn clean_html(html: &str, base_url: &Url) -> Result<String, FullTextParserError> {
@@ -12,34 +13,26 @@ pub fn clean_html(html: &str, base_url: &Url) -> Result<String, FullTextParserEr
     let xpath_ctx = FullTextParser::get_xpath_ctx(&document)?;
     FullTextParser::prep_content(&xpath_ctx, None, &empty_config, base_url, &document);
     if let Some(mut root) = document.get_root_element() {
-       FullTextParser::post_process_page(&mut root)?;
+        FullTextParser::post_process_page(&mut root)?;
     }
     FullTextParser::post_process_document(&document)?;
 
-    // serialize content
-    let options = SaveOptions {
-        format: true,
-        no_declaration: false,
-        no_empty_tags: true,
-        no_xhtml: false,
-        xhtml: false,
-        as_xml: false,
-        as_html: true,
-        non_significant_whitespace: false,
-    };
+    let mut article_node =
+        Node::new("article", None, &document).map_err(|()| FullTextParserError::Xml)?;
+    let content_nodes = Util::evaluate_xpath(&xpath_ctx, "//body/*", true)?;
 
-    if let Some(root) = document.get_root_element() {
-        Ok(document.node_to_string(&root))
-    } else {
-        Ok(document.to_string_with_options(options))
+    for mut node in content_nodes {
+        node.unlink();
+        article_node.add_child(&mut node).unwrap();
     }
+
+    Ok(document.node_to_string(&article_node))
 }
 
 #[cfg(test)]
 mod tests {
-    use reqwest::Url;
     use super::clean_html;
-
+    use reqwest::Url;
 
     #[test]
     fn clean() {
@@ -47,7 +40,6 @@ mod tests {
         let url = Url::parse("https://finshots.in").unwrap();
         let res = clean_html(&html, &url).unwrap();
 
-        println!("{res}");
-        assert_eq!(res.len(), 12118);
+        assert_eq!(res.len(), 11965);
     }
 }
diff --git a/article_scraper/src/full_text_parser/mod.rs b/article_scraper/src/full_text_parser/mod.rs
index 7ae42f6..1f922b5 100644
--- a/article_scraper/src/full_text_parser/mod.rs
+++ b/article_scraper/src/full_text_parser/mod.rs
@@ -58,23 +58,30 @@ impl FullTextParser {
             date: None,
             thumbnail_url: None,
             document: None,
+            root_node: None,
         };
 
-        let mut document = Document::new().map_err(|()| FullTextParserError::Xml)?;
+        let mut new_document = Document::new().map_err(|()| FullTextParserError::Xml)?;
         let mut root =
-            Node::new("article", None, &document).map_err(|()| FullTextParserError::Xml)?;
-        document.set_root_element(&root);
+            Node::new("article", None, &new_document).map_err(|()| FullTextParserError::Xml)?;
+        new_document.set_root_element(&root);
 
-        Self::generate_head(&mut root, &document)?;
+        Self::generate_head(&mut root, &new_document)?;
 
-        let document = Self::parse_html(html, config, global_config)?;
-        let xpath_ctx = Self::get_xpath_ctx(&document)?;
+        let old_document = Self::parse_html(html, config, global_config)?;
+        let xpath_ctx = Self::get_xpath_ctx(&old_document)?;
 
         metadata::extract(&xpath_ctx, config, Some(global_config), &mut article);
         if article.thumbnail_url.is_none() {
             Self::check_for_thumbnail(&xpath_ctx, &mut article);
         }
-        Self::prep_content(&xpath_ctx, config, global_config, &article.url, &document);
+        Self::prep_content(
+            &xpath_ctx,
+            config,
+            global_config,
+            &article.url,
+            &old_document,
+        );
         let found_body = Self::extract_body(&xpath_ctx, &mut root, config, global_config)?;
         if !found_body {
             log::error!("Ftr failed to find content");
@@ -90,9 +97,10 @@ impl FullTextParser {
             return Err(error);
         }
 
-        Self::post_process_document(&document)?;
+        Self::post_process_document(&new_document)?;
 
-        article.document = Some(document);
+        article.document = Some(new_document);
+        article.root_node = Some(root);
         let html = article.get_content().ok_or(FullTextParserError::Scrape)?;
         Ok(html)
     }
@@ -136,6 +144,7 @@ impl FullTextParser {
             date: None,
             thumbnail_url: None,
             document: None,
+            root_node: None,
         };
 
         let mut document = Document::new().map_err(|()| FullTextParserError::Xml)?;
@@ -181,6 +190,7 @@ impl FullTextParser {
         Self::post_process_document(&document)?;
 
         article.document = Some(document);
+        article.root_node = Some(root);
 
         Ok(article)
     }
diff --git a/article_scraper/src/full_text_parser/readability/mod.rs b/article_scraper/src/full_text_parser/readability/mod.rs
index 2de2d85..566f94b 100644
--- a/article_scraper/src/full_text_parser/readability/mod.rs
+++ b/article_scraper/src/full_text_parser/readability/mod.rs
@@ -34,6 +34,7 @@ impl Readability {
             date: None,
             thumbnail_url: None,
             document: None,
+            root_node: None,
         };
 
         let mut article_document = Document::new().map_err(|()| FullTextParserError::Xml)?;
@@ -46,6 +47,7 @@ impl Readability {
         crate::FullTextParser::post_process_document(&article_document)?;
 
         article.document = Some(article_document);
+        article.root_node = Some(root);
         let html = article
             .get_content()
             .ok_or(FullTextParserError::Readability)?;
diff --git a/article_scraper/src/full_text_parser/readability/tests.rs b/article_scraper/src/full_text_parser/readability/tests.rs
index 27aa2c2..f2a58eb 100644
--- a/article_scraper/src/full_text_parser/readability/tests.rs
+++ b/article_scraper/src/full_text_parser/readability/tests.rs
@@ -30,6 +30,7 @@ async fn run_test(name: &str) {
         date: None,
         thumbnail_url: None,
         document: None,
+        root_node: None,
     };
 
     let mut article_document = Document::new().unwrap();
@@ -41,6 +42,7 @@ async fn run_test(name: &str) {
     crate::FullTextParser::post_process_document(&article_document).unwrap();
 
     article.document = Some(article_document);
+    article.root_node = Some(root);
     let html = article.get_content().unwrap();
 
     let expected = std::fs::read_to_string(format!(