pdf txt extraction

cgi/autodoc.fcgi

@@ -6,6 +6,7 @@ use JSON;
 use DBI;
 use GD::Simple;
 use Data::Dumper;
+use Data::UUID;
 use warnings;
 
 $Data::Dumper::Sortkeys = 1;
@@ -61,7 +62,7 @@ while($request->Accept() >= 0) {
 
 	my($code, $hdr, $type, $data) = process_query($method, $path, $qs, $post, $user);
 
-	print STDERR Dumper("code",$code,"hdr", $hdr, "type", $type, "data", $data);
+	#print STDERR Dumper("code",$code,"hdr", $hdr, "type", $type, "data", $data);
 
 	$hdr = [ ] if !defined $hdr;
 
@@ -74,6 +75,11 @@ while($request->Accept() >= 0) {
 	send_response($code, $hdr, $type, $data);
 }
 
+sub gen_uuid {
+	my $ug = Data::UUID->new;
+	return lc($ug->create_str());
+}
+
 sub send_response {
 	my ($code, $hdr, $type, $data) = @_;
 
@@ -254,7 +260,7 @@ sub get_page_image {
 				$size,
 				$dst
 			);
-			print STDERR "CMD=$cmd\n";
+			#print STDERR "CMD=$cmd\n";
 			system($cmd);
 		}
 
@@ -278,16 +284,15 @@ sub get_page_image {
 sub api_v1_POST_documents {
 	my($id, $qs, $post, $user) = @_;
 
+	$id = gen_uuid();
+
 	my $q = sqlquery($dbh, "
 		INSERT INTO documents
 		SET
+			id = ?,
 			owner = ?,
 			status = 'nodata'
-		", $user);
-	$q = sqlquery($dbh, "SELECT LAST_INSERT_ID()");
-	while(my($lastid) = $q->fetchrow_array()) {
-		$id = $lastid;
-	}
+		", $id, $user);
 
 	return db_get_document_object($id);
 }
@@ -295,24 +300,15 @@ sub api_v1_POST_documents {
 sub api_v1_POST_documents_id_data {
 	my($id, $qs, $post, $user) = @_;
 
-	my $pageid;
+	my $uuid = gen_uuid();
+	my $ext = $post->{ctype};
+	$ext =~ s/.*\///;
 
-	sqlquery($dbh, "
-		INSERT INTO
-			pages
-		SET
-			owner = ?,
-			documentId = ?,
-			contenttype = ?,
-			created = NOW(),
-			status = 'inprogress'",
-		$user, $id, $post->{ctype});
-	my $q = sqlquery($dbh, "SELECT LAST_INSERT_ID()");
-	while(my($last) = $q->fetchrow_array()) {
-		$pageid = $last;
-	}
+	my $file = sprintf("%s/%s/%s_%s.%s",
+		$conf->{path}{global},
+		$conf->{path}{queue},
+		$id, $uuid, $ext);
 
-	my $file = $conf->{path}{global} . '/' . $conf->{path}{originals} . '/' . $pageid;
 	open(F,'>'.$file);
 	print F $post->{data};
 	close(F);
@@ -496,7 +492,7 @@ sub sqlquery {
 	my $query = shift;
 	my @args  = @_;
 
-	print STDERR "$query\n";
+	#print STDERR "$query\n";
 
 	my $sth = $dbh->prepare($query) || fatal_api_error(500,"Failed to execute SQL query");
 	$sth->execute(@args) || fatal_api_error(500,"Failed to execute SQL query");

etc/autodoc.json.default

@@ -2,6 +2,7 @@
 	"path": {
 		"global":    "/opt/autodoc",
 		"original":  "var/original",
+		"queue":     "var/queue",
 		"images":    "var/images",
 		"cache":     "var/cache",
 		"error_img": "var/error_img.jpeg"

src/commands.txt

@@ -1,5 +1,7 @@
 apt-get install aspell-fr aspell-it aspell-de aspell-en
+apt-get install wfrench wbritish-large witalian wswiss wngerman wamerican-large
 apt-get install tesseract-ocr-fra tesseract-ocr-deu tesseract-ocr-ita tesseract-ocr-eng
+apt-get install poppler-utils
 
 cd /opt/autodoc/www/js/
 npm install popper.js --save