From 3cd9138c9d8fb9be248c75330cca9da47f4d2e8a Mon Sep 17 00:00:00 2001
From: Lukasz Kasprzak <lukas@labunix.xyz>
Date: Sun, 10 May 2026 22:57:02 +0200
Subject: added a whole bunch of scripts

---
 pdf-txt | 80 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 80 insertions(+)
 create mode 100755 pdf-txt

(limited to 'pdf-txt')

diff --git a/pdf-txt b/pdf-txt
new file mode 100755
index 0000000..92da697
--- /dev/null
+++ b/pdf-txt
@@ -0,0 +1,80 @@
+#!/usr/bin/env bash
+# pdf-txt — extract clean text from OCR'd PDFs, recursively
+#
+# Usage:  pdf-txt <path> [<path>...]
+#
+#   <path> can be a .pdf file or a directory (walked recursively).
+#   For each PDF with a text layer, writes a sibling .txt file using
+#   pdftotext -layout, with "=== Page N ===" markers inserted.
+#   Skips PDFs without a text layer.
+#   Skips PDFs whose .txt is newer than the source.
+
+set -eu
+
+die()  { printf "ERROR: %s\n" "$*" >&2; exit 1; }
+need() { command -v "$1" >/dev/null 2>&1 || die "Missing: $1 (apt install $2)"; }
+
+need pdftotext poppler-utils
+need find       findutils
+need awk        gawk
+
+processed=0; skipped_no_ocr=0; skipped_existing=0; failed=0
+
+extract_one() {
+    local pdf="$1"
+    local txt="${pdf%.[Pp][Dd][Ff]}.txt"
+
+    if [ -f "$txt" ] && [ "$txt" -nt "$pdf" ]; then
+        printf "  up-to-date  %s\n" "$pdf"
+        skipped_existing=$((skipped_existing + 1))
+        return 0
+    fi
+
+    # OCR/text-layer detection: sample first 3 pages
+    local tmp; tmp=$(mktemp)
+    pdftotext -l 3 "$pdf" "$tmp" 2>/dev/null || true
+    local nchars; nchars=$(tr -d '[:space:]' < "$tmp" | wc -c)
+    rm -f "$tmp"
+
+    if [ "$nchars" -lt 50 ]; then
+        printf "  no OCR      %s\n" "$pdf"
+        skipped_no_ocr=$((skipped_no_ocr + 1))
+        return 0
+    fi
+
+    if pdftotext -layout "$pdf" - 2>/dev/null | awk '
+        BEGIN { page = 1; printf "=== Page %d ===\n\n", page }
+        /\014/ { page++; printf "\n=== Page %d ===\n\n", page; next }
+        { print }
+    ' > "$txt"; then
+        printf "  extracted   %s\n" "$txt"
+        processed=$((processed + 1))
+    else
+        rm -f "$txt"
+        printf "  FAILED      %s\n" "$pdf" >&2
+        failed=$((failed + 1))
+    fi
+}
+
+walk() {
+    local target="$1"
+    if [ -f "$target" ]; then
+        case "$target" in
+            *.pdf|*.PDF) extract_one "$target" ;;
+            *) printf "  not a PDF   %s\n" "$target" >&2 ;;
+        esac
+    elif [ -d "$target" ]; then
+        while IFS= read -r -d '' pdf; do
+            extract_one "$pdf"
+        done < <(find "$target" -type f -iname '*.pdf' -print0 | sort -z)
+    else
+        printf "  not found   %s\n" "$target" >&2
+    fi
+}
+
+[ $# -ge 1 ] || die "Usage: pdf-txt <path> [<path>...]"
+
+for arg in "$@"; do walk "$arg"; done
+
+printf "\nDone.  processed=%d  no_ocr=%d  up_to_date=%d  failed=%d\n" \
+    "$processed" "$skipped_no_ocr" "$skipped_existing" "$failed"
-- 
cgit v1.3