Here is the macro for this task using two complex
Perl regular expression Replace All using back-references.
InsertMode
ColumnModeOff
HexOff
Top
PerlReOn
Find MatchCase RegExp "
(?s)(<disp-formula id="deqn
)[^"]*?
("(?:.(?!/disp-formula))+?.\\tag\{
)([^}]+?
)(\}(?:.(?!/disp-formula))+.\\tag\{
)([^}]+?
)\}"
Replace All "
\1\3-
\5\2\3\4\5}"
Top
Find MatchCase RegExp "
(?s)(<disp-formula id="deqn
)[^"]*?
("(?:.(?!/disp-formula|\\tag))+?.\\tag\{
)([^}]+?
)(\}(?:.(?!/disp-formula|\\tag))+?</disp-formula>
)"
Replace All "
\1\3\2\3\4"
Explanation for first search string:
(?s) ... this expression at beginning of search string results in
. matching also newline characters which is by default not the case.
(<disp-formula id="deqn
) ... begin of string to find enclosed in marking (capturing) parentheses to avoid the need to write the same string once again in replace string and instead back-reference this fixed part of found string with
\1 in replace string.
[^"]*? ...
0 or more characters NOT being a double quote non greedy because of question mark after asterisk multiplier. The non greedy matching behavior results in stopping matching any character except double quote on first double quote found even if that results in a negative match for the entire expression.
(...
) ... second marking group to keep the string matched by the expression inside the parentheses unmodified in replace string back-referenced with
\2.
"
(?:.(?!/disp-formula))+?
\\tag
\{ ... this expression matches
- a double quote,
- any character (including newline characters) 1 or more times non greedy because of question mark after plus,
- any character and \tag{ whereby the backslash and the opening brace must be both escaped with a backslash to be interpreted as literal characters.
The special here is that not simply
.+? is used to match any character up to the character left of next
\tag{ as this would also match beyond current
disp-formula element if it does not contain
\tag{ at all, but there is one anywhere in file below.
A character is only matched if the string right of it is not end tag of
disp-formula element.
(?!/disp-formula) is a negative lookahead which does not match characters. It just checks if next to current character there is NOT the string
/disp-formula otherwise the current character is not matched which would be the left angle bracket.
A non marking group
(?:...
) is used around expression
.(?!/disp-formula) to be able to apply the plus multiplier for matching
1 or more characters where next there is NOT the string
/disp-formula.
The expression so far matches everything from start tag of
disp-formula up to first
\tag{ found within same element.
([^}]+?
) ... this expression in third marking group matches
1 or more characters not being a closing brace non greedy.
(\}
(?:.(?!/disp-formula)) \\tag
\{
) ... the expression in fourth capturing group is nearly the same as the expression in second capturing group. It starts with an escaped closing brace instead of a double quote. But the main difference is the missing question mark after plus multiplier. Therefore the matching behavior for
. is now greedy. The matching does not stop on next
\tag{ found within same
disp-formula element because of greedy matching behavior. Now the expression matches as much characters as possible to get nevertheless a positive match for the entire search expression. Therefore this time
. matches everything up to last
\tag{ in same
disp-formula element.
([^}]+?
)\} ... this expression matches
1 or more characters not being a closing brace non greedy in fifth marking group and the closing brace which is specified also in replace string to be kept unmodified.
The replace string of first search string keeps most of found string unmodified because of back-referencing (nearly) everything found in right order with the exception of the
deqn identifier string which is modified with string of first and last
\tag{...
} separated by a hyphen.
The first search string matches only
disp-formula elements with
at least two \tag{...
} embedded.
The second search string is for
disp-formula elements with
exactly one \tag{...
} embedded.
The second search expression is similar to first one. The main difference are the two negative lookaheads containing now not only
/disp-formula, but on OR expression with
/disp-formula OR
\tag.
When first special defined
.+? expression with a negative lookahead in a non marking group stops matching any character, the string right of last matched character must be any character and
\tag{.
And when second special defined
.+? expression with a negative same lookahead in a non marking group stops matching any character, the string right of last matched character must be end tag of same
disp-formula element.
These two restrictions make sure that the second search does not match
- more than one disp-formula element,
- a disp-formula element with no \tag{...} embedded,
- a disp-formula element with more than one \tag{...} embedded.
