Reading a compressed PDF with no PDF tools: inflate the FlateDecode streams with zlib and pull the Tj/TJ strings

Situation: a web fetch tool returned a PDF as raw binary ("FlateDecode streams, text not readable") and saved it to disk; the machine had no pdftotext/poppler and no PyPDF installed. Needed to search the document for a few rules.

Worked in plain Python stdlib (good enough for text-based PDFs, not scans):

import re, zlib
b = open("doc.pdf","rb").read()
out = []
for m in re.finditer(rb"stream\r?\n(.*?)\r?\nendstream", b, re.S):
    try: out.append(zlib.decompress(m.group(1)).decode("latin1"))
    except Exception: pass   # images, fonts, non-Flate streams
txt = " ".join(out)
parts = re.findall(r"\((.*?)\)\s*Tj|\[(.*?)\]\s*TJ", txt)
text = " ".join(a if a else "".join(re.findall(r"\((.*?)\)", t)) for a, t in parts)

Then regex-search text with context windows. Caveats: word spacing from TJ kerning arrays can be lost or doubled, fonts with custom encodings (Identity-H / CID) give garbage and need a real parser, and object streams/xref streams are skipped harmlessly. Run Python with -I on downloaded files and keep the file in its own folder.