Taglists

Grzegorz Kraszewski

---

This article in other languages: Polish

What Is a Taglist?

A taglist is an array of "key-value" pairs. The key is always a 32-bit integer number and is called a tag. The value also has a size of 32 bits. It may be an integer, or a pointer to any structure or object. Taglists are commonly used in the MorphOS API for passing a variable number of arguments, usually sets of attributes with their values. A few special key values are used for array termination, concatenation and item skipping. A set of functions in the utility.library may be used for taglist traversing, filtering, searching, copying etc.

Every pair in a taglist is a TagItem structure, defined in <utility/tagitem.h>:

struct TagItem
{
  ULONG ti_Tag;
  ULONG ti_Data;
};

To be self descriptive, every taglist, being just a plain C array, must have some kind of termination. It is very similar to the string null-termination idea. The termination is done with a TagItem having its ti_Tag field set to TAG_END (which happens to be defined as zero). The ti_Data value of the terminating TagItem is ignored, it is usually set to zero too. The illustration below shows a simple taglist:

This taglist may be created with the following code:

double x = 3.438872763e+17;
Object *obj = NewObject( /* ... */ );

struct TagItem mytaglist[] = {
  { Tag1, 2837 },
  { Tag2, (ULONG)&x },
  { Tag3, (ULONG)"The quick brown fox..." },
  { Tag4, (ULONG)obj },
  { TAG_END, 0 }
};

Passing Taglists to Functions

Building taglists as global or local variables is not very convenient. That is why almost every MorphOS API function getting a taglist as one of its arguments has two forms. The first one accepts a pointer to the taglist. The second one is a variadic args macro building the taglist on the program stack dynamically. Such function pairs are named according to one of the two conventions:

1. SomeFunctionA() takes a pointer to a taglist, SomeFunction() builds the taglist dynamically.
2. SomeFunctionTagList() takes a pointer to a taglist, SomeFunctionTags() builds the taglist dynamically.

Continuing the above example, one can pass the example taglist to SomeFunction() in two ways:

SomeFunctionTagList(mytaglist);
SomeFunctionTags(Tag1, 2837, Tag2, (ULONG)&x, Tag3, (ULONG)"The quick brown fox...", Tag4,
 (ULONG)obj, TAG_END);

Of course in the second case variable mytaglist need not be defined anywhere. Note that for every taglist-based function the taglist is the last argument. There may be some plain arguments before it. It is a common practice to omit ti_Data for the terminator (it is ignored anyway).

Special Tags

Special tags are used to avoid copying large blocks of data when taglists are manipulated. As taglists are plain arrays, operations like removing or inserting elements or merging taglists involve copying their large fragments. Special tags allow for doing these operations by only changing a few tags. Of course this comes at a price. Taglists with special tags (other than TAG_END) may be only manipulated with a set of functions from utility.library. All these functions detect special tags and interpret them properly. It should be noted, that even if taglists are created by hand and do not contain special tags, such tags may appear later, when taglists are passed to system functions. It is therefore recommended to always use utility.library tools to manipulate taglists, especially the taglist iterator named NextTagItem().

TAG_END

I've mentioned this tag already, it works as a taglist terminator. Its value is 0.

TAG_IGNORE

This special tag is used to logically remove a tag item from a taglist without copying large blocks of data. When a tag is to be removed, it is replaced with TAG_IGNORE, tags after it need not to be moved. Any taglist processing function will ignore the tag and advance to the next one immediately.

TAG_MORE

This tag is used for joining taglists. Joining of plain arrays would require counting their size, allocating memory area and copying all tagitems. It is much simplier with TAG_MORE. If one wants to append taglist B at the end of taglist A, he replaces the terminator of taglist A with TAG_MORE and sets the data field to the address of taglist B. Then, when any taglist manipulation function sees this special tag, it fetches the next tag from the address. In effect taglists are logically joined into one.

TAG_SKIP

This rarely used tag is used to logically remove larger fragments from a taglist. When encountered, it directs the taglist iterator to skip a specified number of following tags. The number is specified in the data field of TAG_SKIP.

When using TAG_SKIP one should take care to not skip past the taglist terminator. Taglist processing functions are not protected against such a bug and will try to process random data. Results are unpredictable.

Traversing Taglists With NextTagItem()

NextTagItem() function from the utlilty.library is the basic taglist iterator. It should be always used to traverse taglists, as it supports all the special tags discussed above. Every call to the function returns an address of the next item in the taglist, of course special items are "executed" and not returned. The argument of the function is an address of variable being a pointer to the current tag. This pointer is initialized manually to the start of taglist and should not be modified later. A basic loop for taglist browsing is organized as follows:

struct TagItem *tag, *tagptr;

tagptr = my_taglist;  // initialization to the start of taglist to be traversed

while (tag = NextTagItem(&tagptr))
{
  /* do something with 'tag' */
}

NextTagItem() returns NULL when it encounters TAG_END terminator.

Taglists Processing

Any processing of taglists can be done with just NextTagItem() and manual manipulation of tagitems. On the other hand utility.library provides set of functions for typical operations on taglists. Using them saves time and avoids bugs, which may be possibly made in manually written code. Advanced taglist processing is usually not needed in typical aplications, but comes handy, when taglists are used for data storage.

It should be noted, that this article does not replace autodocs of utility.library functions. These autodocs have been completely rewritten and heavily extended in the MorphOS SDK (compared to Amiga ones). This text is just an overview of possibilities, while autodocs provide detailed descriptions.

Finding Tags and Data

Finding particular tag in a taglist is relatively simple task. It can be done with FindTagItem() function. It returns the first occurence of specified tag. If a tag is expected to appear multiple times, the function may be called in a loop, taking its result as an argument for the next call.

struct TagItem *tag = my_taglist;

while (tag = FindTagItem(SOME_TAG, tag))
{
  /* do something with 'tag' */

  tag++;  /* start from the next tag after the one just found */
}

As all other functions described below, FindTagItem() automatically interprets special tags, so they are never returned to the caller. That is why incrementing the tag pointer in the above loop is always safe.

In many cases (like constructors of MUI classes) we are interested in data of particular tag, and also want a default value if the tag is not present in a taglist. These two operations may be merged into one call to GetTagData() function. It takes a tag, the default value and a taglist as arguments. If the tag is found, its data is returned, if not the result is the default value. Here is an example, the code:

if (tag = FindTagItem(SOME_TAG, taglist)) value = tag->ti_Data; 
else value = DEF_VALUE;

reduces to

value = GetTagData(SOME_TAG, DEF_VALUE, taglist);

The last function related to finding is TagInArray(). It does not operate on a taglist however. It simply checks if a tag is found in a zero-terminated array of tags. This is not a taglist, just plain array of ULONG-s.

Creation and Copying

The simplest way of creating a tagilst is to declare it as an array, be it global or local variable. It has been shown already in this article. If a taglist should be allocated dynamically, one can use any general purpose memory allocation function provided by exec.library. An example with AllocVec():

struct TagItem *taglist;    /* taglist with 7 items (including terminator) */

taglist = (struct TagItem*)AllocVec(7 * sizeof(struct TagItem), MEMF_ANY);

The utility.library provides also a special function for taglist allocation named AllocateTagItems(), which is in fact nothing more than a wrapper on general purpose memory allocation function. It also clears allocated memory block, so it can be treated as an empty taglist (as TAG_END is 0). Taglist allocated with AllocateTagItems() must be freed by FreeTagItems().

A copy of an existing taglist can be created with CloneTagItems(). This function is not a simple wrapper, as taglist containing special tags cannot be copied as a memory block. For doing a proper copy, the function traverses the taglist (using NextTagItem()) two times. First it counts plain tags. Then it allocates a memory block for a clone and does second pass over original, copying tags one by one. Order of tags is preserved, so when we iterate both original and clone with NextTagItem(), we get the same results. The difference is that the clone is stripped off of all special tags. All tags ignored and skipped are not cloned, also separate fragments chained with TAG_MORE are merged into one continuous block. The only special tag present in the clone is, of course, TAG_END.