update extractor for pdf 8.0

src/extract.rs

@@ -1,7 +1,7 @@
 use std::{collections::HashMap, sync::Arc};
 
 use lazy_static::lazy_static;
-use pdf::{backend::Backend, content::Operation, primitive::Primitive};
+use pdf::{backend::Backend, content::{Op, Point, TextDrawAdjusted}};
 use regex::Regex;
 use time::Date;
 
@@ -38,7 +38,7 @@ impl DataSet {
             if let Some(text) = doc_iter.next() {
                 if is_new_column_header(&text.text) {
                     let mut column_name = text.text;
-                    let column_x = text.x;
+                    let column_x = text.point.x;
                     while let Some(more) = doc_iter.peek() {
                         if is_new_column_header(&more.text) || DATE_REGEX.is_match(&more.text) {
                             columns.push((Arc::new(column_name), column_x));
@@ -51,7 +51,7 @@ impl DataSet {
                 } else if DATE_REGEX.is_match(&text.text) {
                     break (
                         DataPoint::new(&text).expect("Failed to parse date!"),
-                        text.y,
+                        text.point.y,
                     );
                 }
             } else {
@@ -67,12 +67,12 @@ impl DataSet {
                     &mut current_datapoint,
                     DataPoint::new(&text).expect("Failed to parse date!"),
                 ));
-                current_y = text.y;
+                current_y = text.point.y;
             } else if VALUE_REGEX.is_match(&text.text) {
-                if (current_y - text.y).abs() > POSITION_ERROR_MARGIN {
+                if (current_y - text.point.y).abs() > POSITION_ERROR_MARGIN {
                     continue;
                 }
-                if let Some((column, _)) = columns.iter().rev().find(|(_, x)| *x < text.x) {
+                if let Some((column, _)) = columns.iter().rev().find(|(_, x)| *x < text.point.x) {
                     current_datapoint.values.insert(
                         column.clone(),
                         text.text.parse().expect("Failed to parse value!"),
@@ -103,26 +103,23 @@ impl DataPoint {
 }
 
 struct DocumentText {
-    x: f32,
-    y: f32,
+    point: Point,
     text: String,
 }
 
 struct DocumentIterator<'a> {
-    x: f32,
-    y: f32,
-    operations: Box<dyn Iterator<Item = Operation> + 'a>,
+    point: Point,
+    operations: Box<dyn Iterator<Item = Op> + 'a>,
 }
 
 impl<'a> DocumentIterator<'a> {
     fn new<B: Backend>(document: &'a pdf::file::File<B>) -> Self {
         Self {
-            x: 0.0,
-            y: 0.0,
+            point: Point { x: 0.0, y: 0.0 },
             operations: Box::new(
                 document
                     .pages()
-                    .filter_map(|page| Some(page.ok()?.contents.clone()?.operations.into_iter()))
+                    .filter_map(|page| Some(page.ok()?.contents.clone()?.operations(document).ok()?.into_iter()))
                     .flatten(),
             ),
         }
@@ -133,54 +130,44 @@ impl<'a> Iterator for DocumentIterator<'a> {
     type Item = DocumentText;
 
     fn next(&mut self) -> Option<Self::Item> {
-        for Operation { operator, operands } in self.operations.as_mut() {
-            if operator == "Tm" {
-                if let (Some(x), Some(y)) = (
-                    operands.get(4).and_then(extract_number),
-                    operands.get(5).and_then(extract_number),
-                ) {
-                    self.x = x;
-                    self.y = y;
+        for operation in self.operations.as_mut() {
+            match operation {
+                Op::SetTextMatrix { matrix } => {
+                    self.point = Point { x: matrix.e, y: matrix.f };
                 }
-            } else if operator == "TJ" || operator == "Tj" {
-                if let Some(text) = operands.get(0).and_then(extract_string) {
-                    return Some(DocumentText {
-                        x: self.x,
-                        y: self.y,
-                        text,
-                    });
+                Op::TextDraw { text } => {
+                    if let Ok(text) = text.to_string() {
+                        return Some(DocumentText {
+                            point: self.point,
+                            text,
+                        });
+                    }
                 }
+                Op::TextDrawAdjusted { array } => {
+                    if let Some(text) = concatenate_adjusted_text(array) {
+                        return Some(DocumentText {
+                            point: self.point,
+                            text,
+                        });
+                    }
+                }
+                _ => continue,
             }
         }
         None
     }
 }
 
-fn extract_number(p: &Primitive) -> Option<f32> {
-    match p {
-        Primitive::Number(n) => Some(*n),
-        Primitive::Integer(n) => Some(*n as f32),
-        _ => None,
-    }
-}
-
-fn extract_string(p: &Primitive) -> Option<String> {
-    let result: Box<dyn AsRef<str>> = match p {
-        Primitive::Array(array) => {
-            let mut acc = String::new();
-            for element in array.iter() {
-                if let Primitive::String(s) = element {
-                    acc += s.as_str().ok()?.as_ref();
-                }
-            }
-            Box::new(acc)
+fn concatenate_adjusted_text(array: Vec<TextDrawAdjusted>) -> Option<String> {
+    let mut acc = String::new();
+    for element in array.iter() {
+        if let TextDrawAdjusted::Text(s) = element {
+            acc += s.to_string().ok()?.as_ref();
         }
-        Primitive::String(s) => Box::new(s.as_str().ok()?),
-        _ => return None,
     };
     Some(
         WHITESPACE_REGEX
-            .replace_all((*result).as_ref().trim(), " ")
+            .replace_all(acc.trim(), " ")
             .to_string(),
     )
 }