ixml parser: read a CDATA section

[oblomov-dev]

What is wrong

A document that carries a CDATA section does not parse. <![CDATA[ falls into
the branch for a start or close tag, where the tag regex does not match it:

ELSEIF lv_xml CP '<*'.
* start or close tag
  FIND FIRST OCCURRENCE OF REGEX lc_regex_tag IN lv_xml RESULTS ls_match.
  ASSERT ls_match-offset = 0.

so the parse ends in ASSERTION_FAILED. That is an assertion, not an
exception — no CATCH cx_root around the parse absorbs it, the caller has no
way to handle the document at all:

parse( |<root><item><![CDATA[a<b&c]]></item></root>| ).
" ASSERTION_FAILED

A section is the one form in which < and & may stand in a document as
text, so it is what a producer reaches for whenever a value carries markup —
an embedded XML or HTML fragment, a snippet of source code, a serialized
payload inside a field.

The change

The loop takes a section ahead of the tag branch: everything between
<![CDATA[ and ]]> becomes the value of a #text node, exactly as it
stands. Nothing in it is unescaped — that is what the section is for, an
< inside one is the four characters < and not <.

ELSEIF lv_xml CP '<![CDATA[*'.
  FIND FIRST OCCURRENCE OF ']]>' IN lv_xml MATCH OFFSET lv_offset.
  ASSERT sy-subrc = 0.
  lv_length = lv_offset - c_cdata_length.
  lv_value = lv_xml+c_cdata_length.
  lv_value = lv_value(lv_length).
  ...

A document without a section is not touched — the new branch is only reached
by the literal <![CDATA[.

Tests

• parse_cdata — <![CDATA[a<b&c]]> reads as a<b&c
• parse_cdata_not_unescaped — <![CDATA[a&lt;b]]> stays a<b
• parse_cdata_newline — a line break inside a section survives
• parse_cdata_empty — <![CDATA[]]> is an empty value, not a failure

npm test is green: abaplint: 0 issue(s) found, 735 file(s) analyzed, and
the full unit run passes.

Not in this PR

A comment fails the same way and for the same reason — <!-- c --> reaches
the tag branch and ends in ASSERTION_FAILED (measured on this branch). It
wants its own change: a comment is skipped rather than turned into a node,
and it may sit before the root element, where a section may not.

[larshp]

conflicts

[larshp]

conflicts