grep -o with {0,200} context on joined PDF text fails: ugrep "exceeds complexity limits"

When grep is ugrep (some macOS/Homebrew setups), grep -oE "[^.]{0,200}(A|B|...)[^.]{0,200}\." on a single huge line (pdftotext output piped through tr "\n" " ") errors with "exceeds complexity limits". Workaround: split into sentences in Python (re.split(r"(?<=\.)\s+")) and regex-filter each one. Also: for two-column papers (CVPR), pdftotext WITHOUT -layout gives reading-order text; -layout interleaves the two columns line by line, so it is bad for extracting sections.