Extracting IPTC data from a JPEG file

When using libiptcdata, the first task you are probably interested in is reading the IPTC metadata from an existing file. At present, JPEG is the only format supported by libiptcdata. The API is designed in a multi-level fashion so that if your application already does some form of JPEG decoding, you can integrate libiptcdata without having to parse the file twice.

The IPTC IIM standard was originally conceived so that the JPEG file itself would be encapsulated inside an IPTC wrapper. Since this is not backward compatable with regular JPEGs, Adobe devised its own standard where an IPTC data block is placed inside the Adobe Photoshop application-specific header. This Photoshop metadata is stored in the APP13 header section of a JPEG file. So reading the IPTC data block is a two-step process: First, extract Photoshop metadata from the APP13 header, and second, extract the IPTC data block from the Photoshop metadata.

More information about the history of IPTC metadata is available at controlledvocabulary.com.

For the impatient, the IPTC data can be extracted with a single line of code using the iptc_data_new_from_jpeg() function if you have no need for fine-grained error checking or you don't want to share code with a pre-existing JPEG decoding routine in your application. Here's an example:

IptcData * d;
char * filename = "foo.jpg";

d = iptc_data_new_from_jpeg (filename);
if (!d) {
	fprintf (stderr, "Error reading IPTC data\n");
	return 1;
}
	

However, for all but the most simplistic of applications, you will probably want finer control of the process. For that, you use three functions. The first, iptc_jpeg_read_ps3() extracts the Photoshop metadata from the APP13 header. If your application already decodes the JPEG file, you don't need this function: Simply store the contents of the APP13 header as you decode the JPEG. See libjpeg Interoperability for an example.

The second function, iptc_jpeg_ps3_find_iptc() extracts the IPTC data block from the Photoshop metadata.

The third function, iptc_data_new_from_data() takes the IPTC data block and decodes it into a data structure of type IptcData which you can use with the rest of the libiptcdata API. Here's an example of all three functions in use:

FILE * infile;
int ps3_len, iptc_off, iptc_len;
IptcData * d;
char * filename = "foo.jpg";
unsigned char buf[256*256];

infile = fopen (filename, "r");
if (!infile) {
	fprintf (stderr, "Error opening %s\n", filename);
	return 1;
}

ps3_len = iptc_jpeg_read_ps3 (infile, buf, sizeof(buf));
fclose (infile);
if (ps3_len < 0) {
	fprintf (stderr, "Error parsing JPEG file\n");
	return 1;
}

if (ps3_len == 0) {
	fprintf (stderr, "File contains no PS3 header\n");
	return 1;
}

iptc_off = iptc_jpeg_ps3_find_iptc (buf, ps3_len, &iptc_len);
if (iptc_off < 0) {
	fprintf (stderr, "Error parsing PS3 header\n");
	return 1;
}

if (iptc_off == 0) {
	fprintf (stderr, "PS3 header contains no IPTC data\n");
	return 1;
}

d = iptc_data_new_from_data (buf + iptc_off, iptc_len);
if (!d) {
	fprintf (stderr, "Error parsing IPTC data\n");
	return 1;
}
	

Creating IPTC data from scratch

If you want to create an empty IptcData structure from scratch without reading it from a preexisting file, use the iptc_data_new() function:

IptcData * d;

d = iptc_data_new ();
	

Viewing IPTC data

IPTC metadata is simply a list of "datasets." Each dataset contains a record number, dataset number, and the data itself. The record number and dataset number, taken together, determine the purpose of the dataset, according to the IPTC IIM specification. For example, dataset 2:120 (record 2, dataset 120) contains an image caption. Some datasets can be repeated, such as dataset 2:25, which contains a single image keyword. Each repeated dataset would contain a separate keyword. The IPTC IIM specification determines the content of each dataset, the type (string, byte, short, etc.), and any restrictions such as minimum/maximum length, repeatability, etc.

The following code snippet shows how to iterate through all the datasets in an IptcData object and print their contents. The iptc_tag_get_title() function gets the string name of a dataset from its record and dataset numbers. iptc_dataset_get_format() retrieves the type of the data contained in the dataset. For datasets with integer types, iptc_dataset_get_value() gets the contents of the tag, and for datasets with string types, iptc_dataset_get_data() will do the same. iptc_dataset_get_as_str() formats the data as a string regardless of its type.

int i;

printf("%6.6s %-20.20s %-9.9s %6s  %s\n", "Tag", "Name", "Type",
		"Size", "Value");
printf(" ----- -------------------- --------- ------  -----\n");

for (i=0; i < d->count; i++) {
	IptcDataSet * e = d->datasets[i];
	char buf[256];

	printf("%2d:%03d %-20.20s %-9.9s %6d  ",
			e->record, e->tag,
			iptc_tag_get_title (e->record, e->tag),
			iptc_format_get_name (iptc_dataset_get_format (e)),
			e->size);
	switch (iptc_dataset_get_format (e)) {
		case IPTC_FORMAT_BYTE:
		case IPTC_FORMAT_SHORT:
		case IPTC_FORMAT_LONG:
			printf("%d\n", iptc_dataset_get_value (e));
			break;
		case IPTC_FORMAT_BINARY:
			iptc_dataset_get_as_str (e, buf, sizeof(buf));
			printf("%s\n", buf);
			break;
		default:
			iptc_dataset_get_data (e, buf, sizeof(buf));
			printf("%s\n", buf);
			break;
	}
}
	

This next code snippet prints out the list of keywords stored in the IPTC metadata. First, in case we don't know the record and dataset numbers for the "Keywords" tag, the iptc_tag_find_by_name() function retrieves them from the IPTC specification. Next, the iptc_data_get_next_dataset() function iterates through each Keyword tag, printing out the contents.

Another important concept presented in this example is the use of reference counts. Whenever a function, such as iptc_data_get_next_dataset() returns a pointer, the reference count for the associated object is incremented by one. It is the application's responsibility to later unreference that pointer so that the count does not keep growing. When the count reaches zero, the object is automatically freed. This must be done for both IptcData (with iptc_data_unref) and IptcDataSet (with iptc_dataset_unref) objects.

IptcDataSet * ds = NULL;
IptcRecord record;
IptcTag tag;
char buf[256];

if (iptc_tag_find_by_name ("Keywords", &record, &tag) < 0) {
	fprintf (stderr, "Invalid tag name\n");
	return 1;
}

while (1) {
	ds = iptc_data_get_next_dataset (d, ds, record, tag);
	if (!ds)
		break;
	iptc_dataset_get_data (ds, buf, sizeof(buf));
	printf("%s\n", buf);
	iptc_dataset_unref(ds);
}
	

Adding a new tag

New tags can be added to an IptcData object with the iptc_data_add_dataset() function. This new dataset can be obtained from another IptcData object, or it can be created from scratch with the iptc_dataset_new() function. If created from scratch, iptc_dataset_set_tag() is first used to set the record and dataset number for the tag. Next, iptc_dataset_set_data() is used to set the actual contents of the tag. The following code snippet demonstrates adding a new keyword, "travel" to an existing set of IPTC data:

IptcDataSet * ds;
IptcRecord record;
IptcTag tag;

if (iptc_tag_find_by_name ("Keywords", &record, &tag) < 0) {
	fprintf (stderr, "Invalid tag name\n");
	return 1;
}

ds = iptc_dataset_new ();
iptc_dataset_set_tag (ds, record, tag);
iptc_dataset_set_data (ds, "travel", 6, IPTC_DONT_VALIDATE);
iptc_data_add_dataset (d, ds);
iptc_dataset_unref (ds);
	

Modifying an existing tag

Combining the functions we have seen above, the following code snippet modifies an existing caption in the IPTC data. It first finds the Caption tag with the iptc_data_get_dataset() function and then changes the data with iptc_dataset_set_data():

IptcDataSet * ds;
IptcRecord record;
IptcTag tag;
char * str = "This is a caption";

if (iptc_tag_find_by_name ("Caption", &record, &tag) < 0) {
	fprintf (stderr, "Invalid tag name\n");
	return 1;
}

ds = iptc_data_get_dataset (d, record, tag);
if (!ds) {
	fprintf (stderr, "Tag not found\n");
	return 1;
}
iptc_dataset_set_data (ds, str, strlen (str), IPTC_DONT_VALIDATE);
iptc_dataset_unref (ds);
	

Deleting a tag

A dataset can be deleted using the iptc_data_remove_dataset() function, as seen in this example, which deletes the Caption from the IPTC data:

IptcDataSet * ds;
IptcRecord record;
IptcTag tag;

if (iptc_tag_find_by_name ("Caption", &record, &tag) < 0) {
	fprintf (stderr, "Invalid tag name\n");
	return 1;
}

ds = iptc_data_get_dataset (d, record, tag);
if (!ds) {
	fprintf (stderr, "Tag not found\n");
	return 1;
}
iptc_data_remove_dataset (d, ds);
iptc_dataset_unref (ds);
	

Saving IPTC data to a JPEG file

Saving an IptcData object back to a JPEG file is a multi-step process requiring several different function calls. Although this may seem cumbersome at first, it allows you to integrate the process into any existing JPEG-saving code already written in your application.

First, a call to iptc_data_sort() sorts the datasets by tag number prior to saving. This step is optional.

Next, by calling iptc_data_save() we convert the IptcData object into a bytestream. This IPTC data block is then encapsulated inside a Photoshop header using iptc_jpeg_ps3_save_iptc(). This function will use an existing Photoshop header, if available, so that all unrelated portions of the Photoshop header are preserved. If no pre-existing Photoshop header is provided, a new one will be created from scratch.

The new Photoshop header can be saved inside the APP13 header of the JPEG file using your application code, or you can use iptc_jpeg_save_with_ps3() to do the same thing.

The following example code demonstrates the entire process:

unsigned char * iptc_buf = NULL;
unsigned char outbuf[256*256];
char tmpfile[strlen(filename)+8];
int v;

iptc_data_sort (d);

if (iptc_data_save (d, &iptc_buf, &iptc_len) < 0) {
	fprintf (stderr, "Failed to generate IPTC bytestream\n");
	return 1;
}
ps3_len = iptc_jpeg_ps3_save_iptc (buf, ps3_len,
		iptc_buf, iptc_len, outbuf, sizeof(outbuf));
iptc_data_free_buf (d, iptc_buf);
if (ps3_len < 0) {
	fprintf (stderr, "Failed to generate PS3 header\n");
	return 1;
}

infile = fopen (filename, "r");
if (!infile) {
	fprintf (stderr, "Can't reopen input file\n");
	return 1;
}
sprintf (tmpfile, "%s.%d", filename, getpid());
outfile = fopen (tmpfile, "w");
if (!outfile) {
	fprintf (stderr, "Can't open temporary file for writing\n");
	return 1;
}

v = iptc_jpeg_save_with_ps3 (infile, outfile, outbuf, ps3_len);
fclose (infile);
fclose (outfile);
if (v < 0) {
	unlink (tmpfile);
	fprintf (stderr, "Failed to save image\n");
	return 1;
}

if (rename (tmpfile, filename) < 0) {
	fprintf (stderr, "Failed to save image\n");
	unlink (tmpfile);
	return 1;
}
fprintf (stderr, "Image saved\n");