chore(page): add box html bytes

spider-rs · Nov 5, 2024 · eb68312 · eb68312
1 parent 9874686
commit eb68312
Show file tree

Hide file tree

Showing 11 changed files with 33 additions and 27 deletions.
diff --git a/Cargo.lock b/Cargo.lock
diff --git a/spider/Cargo.toml b/spider/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "spider"
-version = "2.12.7"
+version = "2.12.8"
 authors = [
     "j-mendez <[email protected]>"
 ]

diff --git a/spider/src/page.rs b/spider/src/page.rs
@@ -123,7 +123,7 @@ pub struct AIResults {
 #[cfg(not(feature = "decentralized"))]
 pub struct Page {
     /// The bytes of the resource.
-    html: Option<Bytes>,
+    html: Option<Box<Bytes>>,
     /// Base absolute url for page.
     base: Option<Url>,
     /// The raw url for the page. Useful since Url::parse adds a trailing slash.
@@ -170,7 +170,7 @@ pub struct Page {
 #[derive(Debug, Clone, Default)]
 pub struct Page {
     /// The bytes of the resource.
-    html: Option<Bytes>,
+    html: Option<Box<Bytes>>,
     #[cfg(feature = "headers")]
     /// The headers of the page request response.
     pub headers: Option<reqwest::header::HeaderMap>,
@@ -318,10 +318,10 @@ pub fn get_page_selectors(url: &str, subdomains: bool, tld: bool) -> Option<Rela
 
 #[cfg(not(feature = "decentralized"))]
 /// Is the resource valid?
-pub fn validate_empty(content: &Option<Bytes>, is_success: bool) -> bool {
+pub fn validate_empty(content: &Option<Box<Bytes>>, is_success: bool) -> bool {
     match content {
         Some(ref content) => {
-            if content.is_empty() || content == "<html><head></head><body></body></html>" || is_success &&
+            if content.is_empty() || content.starts_with(b"<html><head></head><body></body></html>") || is_success &&
             content.starts_with(b"<html>\r\n<head>\r\n<META NAME=\"robots\" CONTENT=\"noindex,nofollow\">\r\n<script src=\"/") && 
             content.ends_with(b"\">\r\n</script>\r\n<body>\r\n</body></html>\r\n") {
                 false
@@ -743,7 +743,10 @@ impl Page {
 
     /// Set the html directly of the page
     pub fn set_html_bytes(&mut self, html: Option<Bytes>) {
-        self.html = html;
+        self.html = match html {
+            Some(html) => Some(Box::new(html)),
+            _ => None,
+        };
     }
 
     /// Set the url directly of the page. Useful for transforming the content and rewriting the url.
@@ -1585,7 +1588,7 @@ pub fn encode_bytes(html: &Bytes, label: &str) -> String {
 
 /// Get the content with proper encoding. Pass in a proper encoding label like SHIFT_JIS.
 #[cfg(feature = "encoding")]
-pub fn get_html_encoded(html: &Option<Bytes>, label: &str) -> String {
+pub fn get_html_encoded(html: &Option<Box<Bytes>>, label: &str) -> String {
     match html.as_ref() {
         Some(html) => encode_bytes(html, label),
         _ => Default::default(),

diff --git a/spider/src/utils/mod.rs b/spider/src/utils/mod.rs
@@ -180,7 +180,7 @@ async fn cf_handle(
 #[derive(Debug, Default)]
 pub struct PageResponse {
     /// The page response resource.
-    pub content: Option<bytes::Bytes>,
+    pub content: Option<Box<bytes::Bytes>>,
     #[cfg(feature = "headers")]
     /// The headers of the response. (Always None if a webdriver protocol is used for fetching.).
     pub headers: Option<reqwest::header::HeaderMap>,
@@ -1402,7 +1402,7 @@ pub async fn handle_response_bytes(
         }
     }
 
-    let mut content: Option<bytes::Bytes> = None;
+    let mut content: Option<Box<bytes::Bytes>> = None;
 
     if !block_streaming {
         let mut stream = res.bytes_stream();
@@ -1433,7 +1433,7 @@ pub async fn handle_response_bytes(
             }
         }
 
-        content.replace(data.into());
+        content.replace(Box::new(data.into()));
     }
 
     PageResponse {

diff --git a/spider_chrome/Cargo.toml b/spider_chrome/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "spider_chrome"
-version = "2.12.7"
+version = "2.12.8"
 rust-version = "1.70"
 authors = [
     "j-mendez <[email protected]>"

diff --git a/spider_cli/Cargo.toml b/spider_cli/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "spider_cli"
-version = "2.12.7"
+version = "2.12.8"
 authors = [
     "j-mendez <[email protected]>"
 ]

diff --git a/spider_transformations/Cargo.toml b/spider_transformations/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "spider_transformations"
-version = "2.12.7"
+version = "2.12.8"
 authors = [
     "j-mendez <[email protected]>"
 ]

diff --git a/spider_transformations/src/html2xml.rs b/spider_transformations/src/html2xml.rs
@@ -4,6 +4,7 @@ use html5ever::{parse_document, QualName};
 use markup5ever::namespace_url;
 use markup5ever::ns;
 use spider::auto_encoder::auto_encode_bytes;
+use spider::bytes::Bytes;
 use spider::page::get_html_encoded;
 use std::default::Default;
 use std::error::Error;
@@ -16,8 +17,10 @@ pub fn convert_html_to_xml(
     encoding: &Option<String>,
 ) -> Result<String, Box<dyn Error>> {
     if encoding.is_some() {
+        let bytes: Box<Bytes> = Box::new(base_convert_xml(html, url, encoding)?.into());
+
         Ok(get_html_encoded(
-            &Some(base_convert_xml(html, url, encoding)?.into()),
+            &Some(bytes),
             &match encoding {
                 Some(encoding) => encoding,
                 _ => "UTF-8",

diff --git a/spider_transformations/src/transformation/mod.rs b/spider_transformations/src/transformation/mod.rs
@@ -50,7 +50,7 @@ mod tests {
         let mut conf = content::TransformConfig::default();
         let mut page_response = PageResponse::default();
 
-        page_response.content = Some(Bytes::from(markup));
+        page_response.content = Some(Bytes::from(markup).into());
         let page = build(url, page_response);
 
         conf.return_format = ReturnFormat::Markdown;
@@ -110,7 +110,7 @@ mod tests {
         let mut conf = content::TransformConfig::default();
         let mut page_response = PageResponse::default();
         conf.return_format = ReturnFormat::XML;
-        page_response.content = Some(Bytes::from(markup));
+        page_response.content = Some(Bytes::from(markup).into());
         let page = build(url, page_response);
         let content = content::transform_content(&page, &conf, &None, &None, &None);
         assert!(
@@ -128,7 +128,7 @@ mod tests {
         let mut conf = content::TransformConfig::default();
         let mut page_response = PageResponse::default();
 
-        page_response.content = Some(Bytes::from(markup));
+        page_response.content = Some(Bytes::from(markup).into());
         let page = build(url, page_response);
 
         conf.return_format = ReturnFormat::Markdown;
@@ -153,7 +153,7 @@ mod tests {
         let mut conf = content::TransformConfig::default();
         let mut page_response = PageResponse::default();
 
-        page_response.content = Some(Bytes::from(markup));
+        page_response.content = Some(Bytes::from(markup).into());
         let page = build(url, page_response);
 
         conf.return_format = ReturnFormat::Markdown;
@@ -183,7 +183,7 @@ mod tests {
         let mut page_response = PageResponse::default();
         let b = Bytes::from(data);
 
-        page_response.content = Some(b);
+        page_response.content = Some(b.into());
 
         let page = build("https://example.com/example.pdf", page_response);
 

diff --git a/spider_utils/Cargo.toml b/spider_utils/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "spider_utils"
-version = "2.12.7"
+version = "2.12.8"
 authors = [
     "j-mendez <[email protected]>"
 ]

diff --git a/spider_worker/Cargo.toml b/spider_worker/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "spider_worker"
-version = "2.12.7"
+version = "2.12.8"
 authors = [
     "j-mendez <[email protected]>"
 ]