✨ now possible to read file contents at any time

Author: ianardee

mindee/http.py

@@ -1,5 +1,5 @@
 import os
-from typing import Optional, Tuple
+from typing import Optional
 
 import requests
 
@@ -82,19 +82,6 @@ def set_api_key_from_env(self) -> None:
             self.api_key = env_key
             logger.debug("Set from environment: %s", self.envvar_key_name)
 
-    @staticmethod
-    def _read_document(
-        input_file: InputDocument, close_file: bool
-    ) -> Tuple[str, bytes]:
-        logger.debug("Reading data from: %s", input_file.filename)
-        input_file.file_object.seek(0)
-        data = input_file.file_object.read()
-        if close_file:
-            input_file.file_object.close()
-        else:
-            input_file.file_object.seek(0)
-        return input_file.filename, data
-
     def predict_request(
         self,
         input_file: InputDocument,
@@ -109,7 +96,7 @@ def predict_request(
         :param close_file: Whether to `close()` the file after parsing it.
         :return: requests response
         """
-        files = {"document": self._read_document(input_file, close_file)}
+        files = {"document": input_file.read_contents(close_file)}
         headers = {"Authorization": self.api_key, "User-Agent": USER_AGENT}
         data = {}
         if include_words:
@@ -132,7 +119,7 @@ def training_request(
         :return: requests response
         :param close_file: Whether to `close()` the file after parsing it.
         """
-        files = {"document": self._read_document(input_file, close_file)}
+        files = {"document": input_file.read_contents(close_file)}
         headers = {"Authorization": self.api_key, "User-Agent": USER_AGENT}
         params = {"training": True, "with_candidates": True}
 

mindee/inputs.py

@@ -2,7 +2,7 @@
 import io
 import os
 from mimetypes import guess_type
-from typing import BinaryIO, Optional
+from typing import BinaryIO, Optional, Tuple
 
 import pikepdf
 
@@ -96,7 +96,7 @@ def is_pdf_empty(self) -> bool:
         """
         Check if the PDF is empty.
 
-        :return: (void) Check if the document contain only empty pages
+        :return: ``True`` if the PDF is empty
         """
         self.file_object.seek(0)
         with pikepdf.open(self.file_object) as pdf:
@@ -131,6 +131,22 @@ def check_pdf_open(self) -> None:
         except Exception as err:
             raise RuntimeError("Couldn't open PDF file") from err
 
+    def read_contents(self, close_file: bool) -> Tuple[str, bytes]:
+        """
+        Read the contents of the input file.
+
+        :param close_file: whether to close the file after reading
+        :return: a Tuple with the file name and binary data
+        """
+        logger.debug("Reading data from: %s", self.filename)
+        self.file_object.seek(0)
+        data = self.file_object.read()
+        if close_file:
+            self.file_object.close()
+        else:
+            self.file_object.seek(0)
+        return self.filename, data
+
 
 class FileDocument(InputDocument):
     def __init__(

requirements.txt

@@ -14,7 +14,7 @@ lxml==4.8.0
     # via pikepdf
 packaging==21.3
     # via pikepdf
-pikepdf==5.0.1
+pikepdf==5.1.2
     # via mindee (setup.py)
 pillow==9.0.1
     # via pikepdf

tests/test_inputs.py

@@ -19,6 +19,17 @@ def test_pdf_reconstruct_ok():
     assert isinstance(input_file.file_object, io.BytesIO)
 
 
+def test_read_contents():
+    input_doc = PathDocument("./tests/data/invoices/invoice.pdf")
+    contents = input_doc.read_contents(close_file=False)
+    assert contents[0] == "invoice.pdf"
+    assert isinstance(contents[1], bytes)
+    assert not input_doc.file_object.closed
+
+    input_doc.read_contents(close_file=True)
+    assert input_doc.file_object.closed
+
+
 def test_pdf_reconstruct_no_cut():
     input_file = PathDocument("./tests/data/invoices/invoice_10p.pdf", cut_pdf=False)
     assert input_file.count_pdf_pages() == 10