Properly support document prefixes (#365)

* Revert "Allow parsing off-spec PDF files with prefixes before the header"

This reverts commit dfb170a.

* Properly support document with prefixing bytes

src/parser/mod.rs

@@ -8,7 +8,7 @@ use std::collections::HashSet;
 use std::str::{self, FromStr};
 
 use nom::branch::alt;
-use nom::bytes::complete::{tag, take, take_until, take_while, take_while1, take_while_m_n};
+use nom::bytes::complete::{tag, take, take_while, take_while1, take_while_m_n};
 use nom::character::complete::multispace1;
 use nom::character::complete::{digit0, digit1, one_of};
 use nom::character::complete::{space0, space1};
@@ -416,16 +416,7 @@ fn _indirect_object<'a>(
 pub fn header(input: ParserInput) -> Option<String> {
     strip_nom(map_res(
         delimited(
-            tuple((
-                map_res(take_until("%PDF-"), |v: ParserInput| {
-                    if v.len() > 1024 {
-                        return Err("Header prefix too long");
-                    }
-
-                    Ok(v)
-                }),
-                tag(b"%PDF-"),
-            )),
+            tag(b"%PDF-"),
             take_while(|c: u8| !b"\r\n".contains(&c)),
             pair(eol, many0_count(comment)),
         ),
@@ -641,22 +632,6 @@ mod tests {
         assert_eq!(real(test_span(b"10.")), Some(10.0));
     }
 
-    #[test]
-    fn parses_malformed_header_with_prefix() {
-        let stream = b"\x20\x20\x20\x20\x0A\x0A\x20\x20%PDF-1.4\n";
-
-        assert_eq!(header(test_span(stream)), Some("1.4".to_string()));
-
-        // 1025 bytes prefix
-        let stream_with_prefix_too_long = [b' '; 1025]
-            .iter()
-            .copied()
-            .chain(b"%PDF-1.4".iter().copied())
-            .collect::<Vec<u8>>();
-
-        assert_eq!(header(test_span(&stream_with_prefix_too_long)), None)
-    }
-
     #[test]
     fn parse_string() {
         let literal_string = |i| tstrip(literal_string(i));

src/reader.rs

@@ -216,6 +216,9 @@ pub const MAX_BRACKET: usize = 100;
 impl Reader<'_> {
     /// Read whole document.
     pub fn read(mut self, filter_func: Option<FilterFunc>) -> Result<Document> {
+        let offset = self.buffer.windows(5).position(|w| w == b"%PDF-").unwrap_or(0);
+        self.buffer = &self.buffer[offset..];
+
         // The document structure can be expressed in PEG as:
         //   document <- header indirect_object* xref trailer xref_start
         let version =
@@ -494,6 +497,15 @@ fn load_short_document() {
     let _doc = Document::load_mem(b"%PDF-1.5\n%%EOF\n").unwrap();
 }
 
+#[test]
+fn load_document_with_preceding_bytes() {
+    let mut content = Vec::new();
+    content.extend(b"garbage");
+    content.extend(include_bytes!("../assets/example.pdf"));
+    let doc = Document::load_mem(&content).unwrap();
+    assert_eq!(doc.version, "1.5");
+}
+
 #[test]
 fn load_many_shallow_brackets() {
     let content: String = std::iter::repeat("()")