Boundary whitespace in direct element constructors

Within a direct element constructor, boundary whitespace is a sequence of consecutive whitespace characters that is delimited at each end either by the start or end of the content, or by a direct constructor, or by an enclosed expression. For example, boundary whitespace might be used in the content of the constructor to separate the end tag of a direct constructor from the start tag of a nested element.

The following diagram shows an example of a direct element constructor, with the boundary whitespace highlighted:

The diagram shows carriage returns and whitespace characters around start and end tags of elements.

The boundary whitespace in this example includes the following characters: a newline character and four space characters that occur between the start tags of the product and description elements; four space characters that occur between the start tag of the description element and the enclosed expression; four space characters that occur between the enclosed expression and the end tag of the description element; and one newline character that appears after the end tag of the description element.

Boundary whitespace does not include any of the following types of whitespace:

Whitespace that is generated by an enclosed expression
Characters that are generated by character references (for example,  ) or by CDataSections
Whitespace characters that are adjacent to a character reference or a CDataSection

The boundary-space policy controls whether boundary whitespace is preserved by element constructors. This policy is specified by a boundary-space declaration in the query prolog. If the boundary-space declaration specifies strip, then boundary whitespace is discarded. If the boundary-space declaration specifies preserve, then boundary whitespace is preserved. If no boundary-space declaration is specified, then the default behavior is to strip boundary whitespace during element construction.

Examples

In the following example, the constructed cat element node has two child element nodes that are named breed and color:
```
<cat>
  <breed>{$b}</breed>
  <color>{$c}</color>
</cat>
```
Because the boundary-space policy is strip by default, the whitespace that surrounds the child elements will be stripped away by the element constructor.
In the following example, the boundary-space policy is strip. This example is equivalent to <a>abc</a>:
```
declare boundary-space strip;
<a> {"abc"} </a>
```
In the following example, however, the boundary-space policy is preserve. This example is equivalent to <a> abc </a>:
```
declare boundary-space preserve;
<a> {"abc"} </a>
```
Because the boundary-space policy is preserve, the spaces that appear before and after the enclosed expression will be preserved by the element constructor.
In the following example, the whitespace that surrounds the z is not boundary whitespace. The whitespace is always preserved, and this example is equivalent to <a> z abc</a>:
```
<a> z {"abc"}</a>
```
In the following example, the whitespace characters that are generated by the character reference and adjacent whitespace characters are preserved, regardless of the boundary-space policy. This example is equivalent to <a> abc</a>:
```
<a>     &#x20;{"abc"}</a>
```
In the following example, the whitespace in the enclosed expression is preserved, regardless of the boundary-space policy, because whitespace that is generated by an enclosed expression is never considered to be boundary whitespace. This example is equivalent to <a> </a>:
```
<a>{" "}</a>
```
The two spaces in the enclosed expression will be preserved by the element constructor and will appear between the start tag and the end tag in the result.