Positional predicate written the long way

A predicate that holds a single number is, by definition, a test on the context position: foo[1] means foo[position() = 1], and foo[last()] means foo[position() = last()]. Spelling the position() = out adds nothing — it is the same selection, just longer to read.

The equality is what makes it redundant, and it counts in either class XPath spells equality in. [position() eq 1] compares two xs:integer values and is true at exactly the position [1] is true at, so the word spelling abbreviates the way the symbol one does. Either operand order reads the same ([1 = position()], [last() eq position()]), and the call is the standard fn:position or fn:last however its namespace is spelled — bare, or behind a prefix bound to the XPath functions namespace. A function of your own that borrows one of those local names is another function and is left alone.

A predicate that asks anything else keeps its position(), which is load-bearing there. [position() = 1 and @current] is a boolean test, and [1 and @current] is a different one. [position() > 1] spans every position but the first rather than naming one. And [position() = '1'] compares against a string, where the abbreviation ['1'] holds a non-empty string and so is true at every position instead of the first.

Incorrect:

<xsl:value-of select="item[position() = 1]"/>
<xsl:apply-templates select="row[position() = last()]"/>
<xsl:value-of select="cell[position() eq last()]"/>

Correct:

<xsl:value-of select="item[1]"/>
<xsl:apply-templates select="row[last()]"/>
<xsl:value-of select="cell[last()]"/>