BUG: PDF size increases because of too high float writing precision (#2213)

See #1910

address regression from #2203

Author: pubpub-zz

pypdf/generic/_base.py

@@ -379,6 +379,9 @@ def readFromStream(
         return IndirectObject.read_from_stream(stream, pdf)
 
 
+FLOAT_WRITE_PRECISION = 8  # shall be min 5 digits max, allow user adj
+
+
 class FloatObject(float, PdfObject):
     def __new__(
         cls, value: Union[str, Any] = "0.0", context: Optional[Any] = None
@@ -409,8 +412,8 @@ def clone(
     def myrepr(self) -> str:
         if self == 0:
             return "0.0"
-        nb = int(log10(abs(self)))
-        s = f"{self:.{max(1,16-nb)}f}".rstrip("0").rstrip(".")
+        nb = FLOAT_WRITE_PRECISION - int(log10(abs(self)))
+        s = f"{self:.{max(1,nb)}f}".rstrip("0").rstrip(".")
         return s
 
     def __repr__(self) -> str:

resources/Seige_of_Vicksburg_Sample_OCR-crazyones-merged.pdf

@@ -1087,7 +1087,9 @@ def test_create_string_object_force():
         ("99900000000000000123", "99900000000000000000"),
         ("99900000000000000123.456000", "99900000000000000000"),
         ("0.00000000000000000000123", "0.00000000000000000000123"),
+        ("0.00000123", "0.00000123"),
         ("0.00000000000000000000123000", "0.00000000000000000000123"),
+        ("-4.6", "-4.6"),  # from #1910
         # (
         #    "50032481330523882508234.00000000000000000000123000",
         #    "50032481330523882508234.00000000000000000000123",

tests/test_generic.py

@@ -985,7 +985,9 @@ def test_colors_in_outline_item(pdf_file_path):
     reader2 = PdfReader(pdf_file_path)
     for outline_item in reader2.outline:
         # convert float to string because of mutability
-        assert [str(c) for c in outline_item.color] == [str(p) for p in purple_rgb]
+        assert ["%.5f" % c for c in outline_item.color] == [
+            "%.5f" % p for p in purple_rgb
+        ]
 
 
 @pytest.mark.samples()
@@ -1619,6 +1621,7 @@ def test_watermark_rendering(tmp_path):
     assert image_similarity(png_path, target_png_path) >= 0.95
 
 
+@pytest.mark.skipif(GHOSTSCRIPT_BINARY is None, reason="Requires Ghostscript")
 def test_watermarking_reportlab_rendering(tmp_path):
     """
     This test is showing a rotated+mirrored watermark in pypdf==3.15.4.