--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/tests/gentle_compare.py	Mon Sep 15 11:44:09 2025 +0200
@@ -0,0 +1,98 @@
+import math
+
+import pymupdf
+
+
+def gentle_compare(w0, w1):
+    """Check lists of "words" extractions for approximate equality.
+
+    * both lists must have same length
+    * word items must contain same word strings
+    * word rectangles must be approximately equal
+    """
+    tolerance = 1e-3  # maximum (Euclidean) norm of difference rectangle
+    word_count = len(w0)  # number of words
+    if word_count != len(w1):
+        print(f"different number of words: {word_count}/{len(w1)}")
+        return False
+    for i in range(word_count):
+        if w0[i][4] != w1[i][4]:  # word strings must be the same
+            print(f"word {i} mismatch")
+            return False
+        r0 = pymupdf.Rect(w0[i][:4])  # rect of first word
+        r1 = pymupdf.Rect(w1[i][:4])  # rect of second word
+        delta = (r1 - r0).norm()  # norm of difference rectangle
+        if delta > tolerance:
+            print(f"word {i}: rectangle mismatch {delta}")
+            return False
+    return True
+
+
+def rms(a, b, verbose=None, out_prefix=''):
+    '''
+    Returns RMS diff of raw bytes of two sequences.
+    '''
+    if verbose is True:
+        verbose = 100000
+    assert len(a) == len(b)
+    e = 0
+    for i, (aa, bb) in enumerate(zip(a, b)):
+        if verbose and (i % verbose == 0):
+            print(f'{out_prefix}rms(): {i=} {e=} {aa=} {aa=}.')
+        e += (aa - bb) ** 2
+    rms = math.sqrt(e / len(a))
+    return rms
+
+
+def pixmaps_rms(a, b, out_prefix=''):
+    '''
+    Returns RMS diff of raw bytes of two pixmaps.
+
+    We assert that the pixmaps/sequences are the same size.
+
+    <a> and <b> can each be a pymupdf.Pixmap or path of a bitmap file.
+    '''
+    if isinstance(a, str):
+        print(f'{out_prefix}pixmaps_rms(): reading pixmap from {a=}.')
+        a = pymupdf.Pixmap(a)
+    if isinstance(b, str):
+        print(f'{out_prefix}pixmaps_rms(): reading pixmap from {b=}.')
+        b = pymupdf.Pixmap(b)
+    assert a.irect == b.irect, f'Differing rects: {a.irect=} {b.irect=}.'
+    a_mv = a.samples_mv
+    b_mv = b.samples_mv
+    assert len(a_mv) == len(b_mv)
+    ret = rms(a_mv, b_mv, verbose=True, out_prefix=out_prefix)
+    print(f'{out_prefix}pixmaps_rms(): {ret=}.')
+    return ret
+
+
+def pixmaps_diff(a, b, out_prefix=''):
+    '''
+    Returns a pymupdf.Pixmap that represents the difference between pixmaps <a>
+    and <b>.
+    
+    Each byte in the returned pixmap is `128 + (b_byte - a_byte) // 2`.
+    '''
+    if isinstance(a, str):
+        print(f'{out_prefix}pixmaps_rms(): reading pixmap from {a=}.')
+        a = pymupdf.Pixmap(a)
+    if isinstance(b, str):
+        print(f'{out_prefix}pixmaps_rms(): reading pixmap from {b=}.')
+        b = pymupdf.Pixmap(b)
+    assert a.irect == b.irect, f'Differing rects: {a.irect=} {b.irect=}.'
+    a_mv = a.samples_mv
+    b_mv = b.samples_mv
+    c = pymupdf.Pixmap(a.tobytes())
+    c_mv = c.samples_mv
+    assert len(a_mv) == len(b_mv) == len(c_mv)
+    if 1:
+        print(f'{len(a_mv)=}')
+        for i, (a_byte, b_byte, c_byte) in enumerate(zip(a_mv, b_mv, c_mv)):
+            assert 0 <= a_byte < 256
+            assert 0 <= b_byte < 256
+            assert 0 <= c_byte < 256
+            # Set byte to 128 plus half the diff so we represent the full
+            # -255..+255 range.
+            c_mv[i] = 128 + (b_byte - a_byte) // 2
+    return c