pdf_read_bytes(bytea) (#2)

README.md

@@ -4,8 +4,37 @@
 
 ## Usage
 
+```sh
+wget https://wiki.postgresql.org/images/e/ea/PostgreSQL_Introduction.pdf -O /tmp/pgintro.pdf
+```
+
+You can use an absolute path to file as a `text` argument
+
+```tsql
+select pdf_read_file('/tmp/pgintro.pdf');
+```
+```tsql
+                                  pdf_read_file                                   
+----------------------------------------------------------------------------------
+ PostgreSQL Introduction                                                         +
+ Digoal.Zhou                                                                     +
+ 7/20/2011Catalog                                                                +
+  PostgreSQL Origin                                                             +
+```
+
+If you don't have the PDF file in your filesystem but have already stored its content in a `bytea` column:
+
 ```tsql
-select pdf_read_file('/abs/path/to/test/pgintro.pdf');
+select pdf_read_bytes(pg_read_binary_file('/tmp/pgintro.pdf'));
+```
+```tsql
+
+                                  pdf_read_bytes                                  
+----------------------------------------------------------------------------------
+ PostgreSQL Introduction                                                         +
+ Digoal.Zhou                                                                     +
+ 7/20/2011Catalog                                                                +
+  PostgreSQL Origin                                                             +
 ```
 
 ## Installation

pgpdf.c

@@ -6,7 +6,7 @@
 #include "poppler/glib/poppler.h"
 
 #if PG_VERSION_NUM >= 160000
-
+#include <glib.h>
 #include "varatt.h"
 
 #endif
@@ -18,38 +18,16 @@
 
 PG_MODULE_MAGIC;
 
-#include <poppler/glib/poppler.h>
-#include <glib.h>
-#include <stdio.h>
-
-
-PG_FUNCTION_INFO_V1(pdf_read_file);
-
-Datum pdf_read_file(PG_FUNCTION_ARGS) {
-    char *pdf_file = text_to_cstring(PG_GETARG_TEXT_PP(0));
-    GError *error = NULL;
-    PopplerDocument *document;
-    gchar *page_text;
-    int num_pages; /* Get the number of pages in the PDF */
-    StringInfo strinfo = makeStringInfo();
-
-    /* Open the PDF document */
-    document = poppler_document_new_from_file(g_strdup_printf("file://%s", pdf_file), NULL, &error);
-    if (!document)
-        elog(ERROR, "Error opening PDF document: %s\n", error->message);
-
-
-    num_pages = poppler_document_get_n_pages(document);
-
-    /* Iterate through pages and extract text */
+static void extract_text_from_pdf(PopplerDocument *document, StringInfo strinfo) {
+    int num_pages = poppler_document_get_n_pages(document);
     for (int i = 0; i < num_pages; i++) {
         PopplerPage *page = poppler_document_get_page(document, i);
         if (!page) {
             elog(WARNING, "Failed to get page %d\n", i);
             continue;
         }
 
-        page_text = poppler_page_get_text(page);
+        gchar *page_text = poppler_page_get_text(page);
         if (page_text) {
             appendStringInfo(strinfo, "%s", page_text);
             g_free(page_text);
@@ -59,8 +37,47 @@ Datum pdf_read_file(PG_FUNCTION_ARGS) {
 
         g_object_unref(page);
     }
+}
+
+static PopplerDocument *open_pdf_file(const char *file_path, GError **error) {
+    return poppler_document_new_from_file(g_strdup_printf("file://%s", file_path), NULL, error);
+}
+
+static PopplerDocument *open_pdf_bytes(bytea *bytes, GError **error) {
+    return poppler_document_new_from_bytes(g_bytes_new(VARDATA_ANY(bytes), VARSIZE_ANY_EXHDR(bytes)), NULL, error);
+}
+
+PG_FUNCTION_INFO_V1(pdf_read_file);
+
+Datum pdf_read_file(PG_FUNCTION_ARGS) {
+    char *pdf_file = text_to_cstring(PG_GETARG_TEXT_PP(0));
+    GError *error = NULL;
+    StringInfo strinfo = makeStringInfo();
+
+    PopplerDocument *document = open_pdf_file(pdf_file, &error);
+    if (!document) {
+        elog(ERROR, "Error opening PDF document: %s\n", error->message);
+    }
+
+    extract_text_from_pdf(document, strinfo);
+    g_object_unref(document);
+
+    PG_RETURN_TEXT_P(cstring_to_text(strinfo->data));
+}
+
+PG_FUNCTION_INFO_V1(pdf_read_bytes);
+
+Datum pdf_read_bytes(PG_FUNCTION_ARGS) {
+    bytea *bytes = PG_GETARG_BYTEA_PP(0);
+    GError *error = NULL;
+    StringInfo strinfo = makeStringInfo();
+
+    PopplerDocument *document = open_pdf_bytes(bytes, &error);
+    if (!document) {
+        elog(ERROR, "Error opening PDF document: %s\n", error->message);
+    }
 
-    /* Close the PDF document */
+    extract_text_from_pdf(document, strinfo);
     g_object_unref(document);
 
     PG_RETURN_TEXT_P(cstring_to_text(strinfo->data));

sql/pgpdf--0.1.0.sql

@@ -1,3 +1,7 @@
 CREATE FUNCTION pdf_read_file(text) returns text AS 'MODULE_PATHNAME',
 'pdf_read_file'
     LANGUAGE C;
+
+CREATE FUNCTION pdf_read_bytes(bytea) returns text AS 'MODULE_PATHNAME',
+'pdf_read_bytes'
+    LANGUAGE C;