ENH: Add level parameter to compress_content_streams (#2044)

Provide more options / details on how to reduce the file size with compression.

See #1910

## A small experiment

```
File    File name
size    The suffix is the compression level
-------------------
5321132 GeoTopo.pdf
9959402 out-0.pdf
5976025 out-1.pdf
5914204 out-2.pdf
5885818 out-3.pdf
5816263 out-4.pdf
5762359 out-5.pdf
5738259 out-6.pdf
5731877 out-7.pdf
5726121 out-8.pdf
5725267 out-9.pdf
```

Level 1 gives a very good improvement, but already level 2 might not be worth the CPU cycles.

Interestingly, the original is smaller than the best compression

docs/user/file-size.md

@@ -1,4 +1,4 @@
-# Reduce PDF Size
+# Reduce PDF File Size
 
 There are multiple ways to reduce the size of a given PDF file. The easiest
 one is to remove content (e.g. images) or pages.
@@ -77,7 +77,8 @@ pypdf supports the FlateDecode filter which uses the zlib/deflate compression
 method. It is a lossless compression, meaning the resulting PDF looks exactly
 the same.
 
-Deflate compression can be applied to a page via [`page.compress_content_streams`](https://pypdf.readthedocs.io/en/latest/modules/PageObject.html#pypdf._page.PageObject.compress_content_streams):
+Deflate compression can be applied to a page via
+[`page.compress_content_streams`](https://pypdf.readthedocs.io/en/latest/modules/PageObject.html#pypdf._page.PageObject.compress_content_streams):
 
 ```python
 from pypdf import PdfReader, PdfWriter
@@ -96,15 +97,29 @@ with open("out.pdf", "wb") as f:
     writer.write(f)
 ```
 
+`page.compress_content_streams` uses [`zlib.compress`](https://docs.python.org/3/library/zlib.html#zlib.compress)
+and supports the `level` paramter: `level=0` means no compression,
+`level=9` refers to the highest compression.
+
 Using this method, we have seen a reduction by 70% (from 11.8 MB to 3.5 MB)
 with a real PDF.
 
 ## Removing Sources
 
-When a page is removed from the page list, its content will still be present in the PDF file. This means that the data may still be used elsewhere.
+When a page is removed from the page list, its content will still be present in
+the PDF file. This means that the data may still be used elsewhere.
 
-Simply removing a page from the page list will reduce the page count but not the file size. In order to exclude the content completely, the pages should not be added to the PDF using the PdfWriter.append() function. Instead, only the desired pages should be selected for inclusion (note: [PR #1843](https://github.com/py-pdf/pypdf/pull/1843) will add a page deletion feature).
+Simply removing a page from the page list will reduce the page count but not the
+file size. In order to exclude the content completely, the pages should not be
+added to the PDF using the PdfWriter.append() function. Instead, only the
+desired pages should be selected for inclusion
+(note: [PR #1843](https://github.com/py-pdf/pypdf/pull/1843) will add a page
+deletion feature).
 
-There can be issues with poor PDF formatting, such as when all pages are linked to the same resource. In such cases, dropping references to specific pages becomes useless because there is only one source for all pages.
+There can be issues with poor PDF formatting, such as when all pages are linked
+to the same resource. In such cases, dropping references to specific pages
+becomes useless because there is only one source for all pages.
 
-Cropping is an ineffective method for reducing the file size because it only adjusts the viewboxes and not the external parts of the source image. Therefore, the content that is no longer visible will still be present in the PDF.
+Cropping is an ineffective method for reducing the file size because it only
+adjusts the viewboxes and not the external parts of the source image. Therefore,
+the content that is no longer visible will still be present in the PDF.

pypdf/_page.py

@@ -1763,7 +1763,7 @@ def scaleTo(self, width: float, height: float) -> None:  # deprecated
         deprecation_with_replacement("scaleTo", "scale_to", "3.0.0")
         self.scale_to(width, height)
 
-    def compress_content_streams(self) -> None:
+    def compress_content_streams(self, level: int = -1) -> None:
         """
         Compress the size of this page by joining all content streams and
         applying a FlateDecode filter.
@@ -1773,7 +1773,7 @@ def compress_content_streams(self) -> None:
         """
         content = self.get_contents()
         if content is not None:
-            content_obj = content.flate_encode()
+            content_obj = content.flate_encode(level)
             try:
                 content.indirect_reference.pdf._objects[  # type: ignore
                     content.indirect_reference.idnum - 1  # type: ignore

pypdf/filters.py

@@ -225,17 +225,18 @@ def _decode_png_prediction(data: str, columns: int, rowlength: int) -> bytes:
         return output.getvalue()
 
     @staticmethod
-    def encode(data: bytes) -> bytes:
+    def encode(data: bytes, level: int = -1) -> bytes:
         """
         Compress the input data using zlib.
 
         Args:
             data: The data to be compressed.
+            level: See https://docs.python.org/3/library/zlib.html#zlib.compress
 
         Returns:
             The compressed data.
         """
-        return zlib.compress(data)
+        return zlib.compress(data, level)
 
 
 class ASCIIHexDecode:

pypdf/generic/_data_structures.py

@@ -880,7 +880,7 @@ def flateEncode(self) -> "EncodedStreamObject":  # deprecated
         deprecation_with_replacement("flateEncode", "flate_encode", "3.0.0")
         return self.flate_encode()
 
-    def flate_encode(self) -> "EncodedStreamObject":
+    def flate_encode(self, level: int = -1) -> "EncodedStreamObject":
         from ..filters import FlateDecode
 
         if SA.FILTER in self:
@@ -909,7 +909,7 @@ def flate_encode(self) -> "EncodedStreamObject":
         retval[NameObject(SA.FILTER)] = f
         if parms is not None:
             retval[NameObject(SA.DECODE_PARMS)] = parms
-        retval._data = FlateDecode.encode(self._data)
+        retval._data = FlateDecode.encode(self._data, level)
         return retval