Add pipeline code

_pipeline/pdftext.py

@@ -0,0 +1,76 @@
+"""Read a PDF's text layer and image positions with poppler. No guessing: coordinates in, text out."""
+import os
+import re
+import subprocess
+import tempfile
+import xml.etree.ElementTree as ET
+
+
+def _run(cmd):
+    return subprocess.run(cmd, check=True, capture_output=True, text=True).stdout
+
+
+def page_geometry(pdf):
+    """[(width, height)] per page as displayed (rotation applied)."""
+    n = int(re.search(r"^Pages:\s+(\d+)", _run(["pdfinfo", pdf]), re.M).group(1))
+    out = _run(["pdfinfo", "-f", "1", "-l", str(n), pdf])
+    sizes = {int(m[0]): (float(m[1]), float(m[2])) for m in re.findall(r"^Page\s+(\d+) size:\s+([\d.]+) x ([\d.]+)", out, re.M)}
+    rots = {int(m[0]): int(m[1]) for m in re.findall(r"^Page\s+(\d+) rot:\s+(\d+)", out, re.M)}
+    geo = []
+    for p in range(1, n + 1):
+        w, h = sizes[p]
+        if rots.get(p, 0) in (90, 270):
+            w, h = h, w
+        geo.append((w, h))
+    return geo
+
+
+def halves(width, height):
+    """A landscape page that holds two book pages is read as two units, left then right."""
+    if width > height * 1.2:
+        return [("L", 0, width / 2), ("R", width / 2, width / 2)]
+    return [(None, 0, width)]
+
+
+def chars_per_page(pdf):
+    """Number of non-space characters in the text layer of each page."""
+    n = len(page_geometry(pdf))
+    out = []
+    for p in range(1, n + 1):
+        txt = _run(["pdftotext", "-f", str(p), "-l", str(p), pdf, "-"])
+        out.append(len(re.sub(r"\s", "", txt)))
+    return out
+
+
+def units(pdf):
+    """Reading units in order: [{'pdf_page', 'half', 'text'}]. Text keeps the layout (indentation, hyphens)."""
+    result = []
+    for p, (w, h) in enumerate(page_geometry(pdf), start=1):
+        for half, x, width in halves(w, h):
+            text = _run(["pdftotext", "-layout", "-f", str(p), "-l", str(p),
+                         "-x", str(int(x)), "-y", "0", "-W", str(int(width)), "-H", str(int(h)), pdf, "-"])
+            result.append({"pdf_page": p, "half": half, "text": text.replace("\f", "")})
+    return result
+
+
+def images(pdf):
+    """Images with their page, half and pixel size: [{'pdf_page', 'half', 'width', 'height'}]."""
+    geo = page_geometry(pdf)
+    found = []
+    with tempfile.TemporaryDirectory() as tmp:
+        subprocess.run(["pdftohtml", "-xml", "-q", pdf, os.path.join(tmp, "o")], check=True, capture_output=True)
+        root = ET.parse(os.path.join(tmp, "o.xml")).getroot()
+        for page in root.iter("page"):
+            p = int(page.get("number"))
+            w, h = geo[p - 1]
+            scale = float(page.get("width")) / w
+            for img in page.iter("image"):
+                cx = (float(img.get("left")) + float(img.get("width")) / 2) / scale
+                half = None
+                for name, x, width in halves(w, h):
+                    if x <= cx < x + width:
+                        half = name
+                found.append({"pdf_page": p, "half": half,
+                              "width": round(float(img.get("width")) / scale),
+                              "height": round(float(img.get("height")) / scale)})
+    return found