One Hat Cyber Team

View File Name : addingtags.html

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<HTML>
<HEAD>
<TITLE>
Modifying The TIFF Library
</TITLE>
<style type="text/css">
body {
font-family: Arial, Helvetica, Sans;
}
</style>
</HEAD>
<BODY>

<H1>
Defining New TIFF Tags
</H1>

Libtiff has built-in knowledge of all the standard TIFF tags, as
well as extensions.  The following describes how to add knowledge of
new tags as builtins to libtiff, or how to application specific tags can
be used by applications without modifying libtiff. 
<p>

<h2>TIFFFieldInfo</h2>

How libtiff manages specific tags is primarily controlled by the 
definition for that tag value stored internally as a TIFFFieldInfo structure. 
This structure looks like this:
<p>

<pre>
typedef	struct {
  ttag_t    field_tag;          /* field's tag */
  short	    field_readcount;    /* read count / TIFF_VARIABLE / TIFF_VARIABLE2 / TIFF_SPP */
  short	    field_writecount;   /* write count / TIFF_VARIABLE / TIFF_VARIABLE2 */
  TIFFDataType field_type;      /* type of associated data */
  unsigned short field_bit;     /* bit in fieldsset bit vector */
  unsigned char field_oktochange;/* if true, can change while writing */
  unsigned char field_passcount;/* if true, pass dir count on set */
  char	*field_name;		/* ASCII name */
} TIFFFieldInfo;
</pre>

<ul>
<li> <b>field_tag</b>: the tag number.  For instance 277 for the
SamplesPerPixel tag.  Builtin tags will generally have a #define in
tiff.h for each known tag. <p>

<li> <b>field_readcount</b>: The number of values which should be read.
The special value TIFF_VARIABLE (-1) indicates that a variable number of
values may be read.  The special value TIFFTAG_SPP (-2) indicates that there
should be one value for each sample as defined by TIFFTAG_SAMPLESPERPIXEL.  
The special value TIFF_VARIABLE2 (-3) is similar to TIFF_VARIABLE
but the required TIFFGetField() count value must be uint32_t* instead of uint16_t* as for TIFF_VARIABLE (-1).  
For ASCII fields with variable length, this field is TIFF_VARIABLE.<p>

<li> <b>field_writecount</b>: The number of values which should be written.
Generally the same as field_readcount.  A few built-in exceptions exist, but
I haven't analysed why they differ. <p>

<li> <b>field_type</b>: Type of the field.  One of TIFF_BYTE, TIFF_ASCII,
TIFF_SHORT, TIFF_LONG, TIFF_RATIONAL, TIFF_SBYTE, TIFF_UNDEFINED, 
TIFF_SSHORT, TIFF_SLONG, TIFF_SRATIONAL, TIFF_FLOAT, TIFF_DOUBLE or
TIFF_IFD.  Note that some fields can support more than one type (for 
instance short and long).  These fields should have multiple TIFFFieldInfos. 
<p>

<li> <b>field_bit</b>: Built-in tags stored in special fields in the 
TIFF structure have assigned field numbers to distinguish them (ie. 
FIELD_SAMPLESPERPIXEL).  New tags should generally just use 
FIELD_CUSTOM indicating they are stored in the generic tag list.<p>

<li> <b>field_oktochange</b>: TRUE if it is OK to change this tag value
while an image is being written.  FALSE for stuff that must be set once
and then left unchanged (like ImageWidth, or PhotometricInterpretation for
instance).<p>

<li> <b>field_passcount</b>: If TRUE, then the count value must be passed
in TIFFSetField(), and TIFFGetField(), otherwise the count is not required.
This should generally be TRUE for non-ascii variable count tags unless
the count is implicit (such as with the colormap).<p>

<li> <b>field_name</b>: A name for the tag.  Normally mixed case (studly caps) 
like "StripByteCounts" and relatively short. <p>

</ul>

A TIFFFieldInfo definition exists for each built-in tag in the tif_dirinfo.c
file.  Some tags which support multiple data types have more than one
definition, one per data type supported. <p>

Various functions exist for getting the internal TIFFFieldInfo definitions,
including _TIFFFindFieldInfo(), and _TIFFFindFieldInfoByName().  See
tif_dirinfo.c for details.  There must be some mechanism to get the whole
list, though I don't see it off hand.<p>

<h2>Default Tag Auto-registration</h2>

In libtiff 3.6.0 a new mechanism was introduced allowing libtiff to 
read unrecognised tags automatically.  When an unknown tags is encountered, 
it is automatically internally defined with a default name and a type 
derived from the tag value in the file.  Applications only need to predefine
application specific tags if they need to be able to set them in a file, or
if particular calling conventions are desired for TIFFSetField() and 
TIFFGetField().<p>

When tags are autodefined like this the <b>field_readcount</b> and
<b>field_writecount</b> values are always TIFF_VARIABLE2 (-3).  The 
<b>field_passcount</b> is always TRUE, and the <b>field_bit</b> is 
FIELD_CUSTOM.  The field name will be "Tag %d" where the %d is the tag 
number.
<BR>Thus, to read anonymous auto-registered tags use the following:
<pre>
    uint32_t count;
    void* value;  //has to be a pointer to the expected value type.
    TIFFGetField(tif, TIFFTAG_UNKNOWN_TO_LIBTIFF, &count, &value);

</pre><p>  

<h2>Defining Application Tags</h2>

For various reasons, it is common for applications to want to define
their own tags to store information outside the core TIFF specification. 
This is done by calling TIFFMergeFieldInfo() with one or more TIFFFieldInfos. 
<p>

The libgeotiff library provides geospatial information extensions within
a TIFF file.  First, a set of TIFFFieldInfo's is prepared with information
on the new tags:<p>

<pre>
static const TIFFFieldInfo xtiffFieldInfo[] = {
  
  /* XXX Insert Your tags here */
    { TIFFTAG_GEOPIXELSCALE,	-1,-1, TIFF_DOUBLE,	FIELD_CUSTOM,
      TRUE,	TRUE,	"GeoPixelScale" },
    { TIFFTAG_GEOTRANSMATRIX,	-1,-1, TIFF_DOUBLE,	FIELD_CUSTOM,
      TRUE,	TRUE,	"GeoTransformationMatrix" },
    { TIFFTAG_GEOTIEPOINTS,	-1,-1, TIFF_DOUBLE,	FIELD_CUSTOM,
      TRUE,	TRUE,	"GeoTiePoints" },
    { TIFFTAG_GEOKEYDIRECTORY, -1,-1, TIFF_SHORT,	FIELD_CUSTOM,
      TRUE,	TRUE,	"GeoKeyDirectory" },
    { TIFFTAG_GEODOUBLEPARAMS,	-1,-1, TIFF_DOUBLE,	FIELD_CUSTOM,
      TRUE,	TRUE,	"GeoDoubleParams" },
    { TIFFTAG_GEOASCIIPARAMS,	-1,-1, TIFF_ASCII,	FIELD_CUSTOM,
      TRUE,	FALSE,	"GeoASCIIParams" }
};
</pre>

In order to define the tags, we call TIFFMergeFieldInfo() on the
desired TIFF handle with the list of TIFFFieldInfos.<p>

<pre>
#define	N(a)	(sizeof (a) / sizeof (a[0]))

    /* Install the extended Tag field info */
    TIFFMergeFieldInfo(tif, xtiffFieldInfo, N(xtiffFieldInfo));
</pre>

The tags need to be defined for each TIFF file opened - and when reading
they should be defined before the tags of the file are read, yet a valid
TIFF * is needed to merge the tags against.   In order to get them 
registered at the appropriate part of the setup process, it is necessary
to register our merge function as an extender callback with libtiff. 
This is done with TIFFSetTagExtender().  We also keep track of the 
previous tag extender (if any) so that we can call it from our extender
allowing a chain of customizations to take effect. <P>

<pre>
static TIFFExtendProc _ParentExtender = NULL;

static
void _XTIFFInitialize(void)
{
    static int first_time=1;
	
    if (! first_time) return; /* Been there. Done that. */
    first_time = 0;
	
    /* Grab the inherited method and install */
    _ParentExtender = TIFFSetTagExtender(_XTIFFDefaultDirectory);
}
</pre>

The extender callback is looks like this.  It merges in our new fields
and then calls the next extender if there is one in effect.<p>

<pre>
static void
_XTIFFDefaultDirectory(TIFF *tif)
{
    /* Install the extended Tag field info */
    TIFFMergeFieldInfo(tif, xtiffFieldInfo, N(xtiffFieldInfo));

    /* Since an XTIFF client module may have overridden
     * the default directory method, we call it now to
     * allow it to set up the rest of its own methods.
     */

    if (_ParentExtender) 
        (*_ParentExtender)(tif);
}
</pre>

The above approach ensures that our new definitions are used when reading
or writing any TIFF file.  However, since on reading we already have 
default definitions for tags, it is usually not critical to pre-define them.
If tag definitions are only required for writing custom tags, you can just
call TIFFMergeFieldInfo() before setting new tags.  The whole extender
architecture can then be avoided.<p>

<H2 id="AddingTags">Adding New Builtin Tags</H2>
<P>
A similar approach is taken to the above.  However, the TIFFFieldInfo 
should be added to the tiffFieldInfo[] list in tif_dirinfo.c.  Ensure that
new tags are added in sorted order by the tag number.<p>

Normally new built-in tags should be defined with FIELD_CUSTOM; however, if
it is desirable for the tag value to have it's own field in the TIFFDirectory
structure, then you will need to #define a new FIELD_ value for it, and
add appropriate handling as follows:


<OL>
<LI>Define the tag in <B>tiff.h</B>.
<LI>Add a field to the directory structure in <B>tif_dir.h</B>
   and define a <TT>FIELD_*</TT> bit (also update the definition of
   <TT>FIELD_CODEC</TT> to reflect your addition).
<LI>Add an entry in the <TT>TIFFFieldInfo</TT> array defined at the top of
   <B>tif_dirinfo.c</B>. 
   Note that you must keep this array sorted by tag
   number and that the widest variant entry for a tag should come
   first (e.g. <TT>LONG</TT> before <TT>SHORT</TT>).
<LI>Add entries in <TT>_TIFFVSetField()</TT> and <TT>_TIFFVGetField()</TT>
   for the new tag.
<LI>(<I>optional</I>) If the value associated with the tag is not a scalar value
   (e.g. the array for <TT>TransferFunction</TT>) and requires
   special processing,
   then add the appropriate code to <TT>TIFFReadDirectory()</TT> and
   <TT>TIFFWriteDirectory()</TT>.  You're best off finding a similar tag and
   cribbing code.
<LI>Add support to <TT>TIFFPrintDirectory()</TT> in <B>tif_print.c</B>
    to print the tag's value.
</OL>

<P>
If you want to maintain portability, beware of making assumptions
about data types.  Use the typedefs (<TT>uint16_t</TT>, etc. when dealing with
data on disk and <TT>t*_t</TT> when stuff is in memory) and be careful about
passing items through printf or similar vararg interfaces.

<H2 id="AddingCODECTags">Adding New Codec-private Tags</H2>
<P>
To add tags that are meaningful <EM>only when a particular compression
algorithm is used</EM> follow these steps:

<OL>
<LI>Define the tag in <B>tiff.h</B>.
<LI>Allocate storage for the tag values in the private state block of
   the codec.
<LI>Insure the state block is created when the codec is initialized.
<LI>At <TT>TIFFInitfoo</TT> time override the method pointers in the
    TIFF structure
   for getting, setting and printing tag values.  For example,
<PRE>
    sp->vgetparent = tif->tif_vgetfield;
    tif->tif_vgetfield = fooVGetField;	/* hook for codec tags */
    sp->vsetparent = tif->tif_vsetfield;
    tif->tif_vsetfield = fooVSetField;	/* hook for codec tags */
    tif->tif_printdir = fooPrintDir;	/* hook for codec tags */
</PRE>
   (Actually you may decide not to override the
   <TT>tif_printdir</TT> method, but rather just specify it).
<LI>Create a private <TT>TIFFFieldInfo</TT> array for your tags and
    merge them into the core tags at initialization time using
    <TT>_TIFFMergeFieldInfo</TT>; e.g.
<PRE>
    _TIFFMergeFieldInfo(tif, fooFieldInfo, N(fooFieldInfo));
</PRE>
   (where <TT>N</TT> is a macro used liberaly throughout the distributed code).
<LI>Fill in the get and set routines.  Be sure to call the parent method
   for tags that you are not handled directly.  Also be sure to set the
   <TT>FIELD_*</TT> bits for tags that are to be written to the file.  Note that
   you can create ``pseudo-tags'' by defining tags that are processed
   exclusively in the get/set routines and never written to file (see
   the handling of <TT>TIFFTAG_FAXMODE</TT> in <B>tif_fax3.c</B>
   for an example of this).
<LI>Fill in the print routine, if appropriate.
</OL>

Note that space has been allocated in the <TT>FIELD_*</TT> bit space for
codec-private tags.  Define your bits as <TT>FIELD_CODEC+<offset></TT> to
keep them away from the core tags.  If you need more tags than there
is room for, just increase <TT>FIELD_SETLONGS</TT> at the top of
<B>tiffiop.h</B>.

<HR>

Last updated: $Date: 2016-09-25 20:05:44 $

</BODY>

</HTML>