Excel - Extract substring(s) from string using FILTERXML

Excel's FILTERXML uses XPATH 1.0 which unfortunately means it is not as diverse as we would maybe want it to be. Also, Excel seems to not allow returning reworked node values and exclusively allows you to select nodes in order of appearance. However there is a fair share of functions we can still utilize. More information about that can be found here.

The function takes two parameters: =FILTERXML(<A string in valid XML format>,<A string in valid XPATH format>)

Let's say cell A1 holds the string: ABC|123|DEF|456|XY-1A|ZY-2F|XY-3F|XY-4f|xyz|123. To create a valid XML string we use SUBSTITUTE to change the delimiter to valid end- and start-tag constructs. So to get a valid XML construct for the given example we could do:

"<t><s>"&SUBSTITUTE(A1,"|","</s><s>")&"</s></t>"

For readability reasons I'll refer to the above construct with the word <XML> as a placeholder. Below you'll find different usefull XPATH functions in a valid construct to filter nodes:

_{1) All Elements:}

=FILTERXML(<XML>,"//s")

_{Returns: ABC, 123, DEF, 456, XY-1A, ZY-2F, XY-3F, XY-4f, xyz and 123 (all nodes)}

_{2) Elements by position:}

=FILTERXML(<XML>,"//s[position()=4]")

Or:

=FILTERXML(<XML>,"//s[4]")

_{Returns: 456 (node on index 4)}

=FILTERXML(<XML>,"//s[position()<4]")

_{Returns: ABC, 123 and DEF (nodes on index < 4)}

=FILTERXML(<XML>,"//s[position()=2 or position()>5]")

_{Returns: 123, ZY-2F, XY-3F, XY-4f, xyz and 123 (nodes on index 2 or > 5)}

=FILTERXML(<XML>,"//s[last()]")

_{Returns: 123 (node on last index)}

=FILTERXML(<XML>,"//s[position() mod 2 = 1]")

_{Returns: ABC, DEF, XY-1A, XY-3F and xyz (odd nodes)}

=FILTERXML(<XML>,"//s[position() mod 2 = 0]")

_{Returns: 123, 456, ZF-2F, XY-4f and 123 (even nodes)}

_{3) (Non) numeric elements:}

=FILTERXML(<XML>,"//s[number()=.]")

Or:

=FILTERXML(<XML>,"//s[.*0=0]")

_{Returns: 123, 456, and 123 (numeric nodes)}

=FILTERXML(<XML>,"//s[not(number()=.)]")

Or:

=FILTERXML(<XML>,"//s[.*0!=0)]")

_{Returns: ABC, DEF, XY-1A, ZY-2F, XY-3F, XY-4f and xyz (non-numeric nodes)}

_{4) Elements that (not) contain:}

=FILTERXML(<XML>,"//s[contains(., 'Y')]")

_{Returns: XY-1A, ZY-2F, XY-3F and XY-4f (containing 'Y', notice XPATH is case sensitive, exclusing xyz)}

=FILTERXML(<XML>,"//s[not(contains(., 'Y'))]")

_{Returns: ABC, 123, DEF, 456, xyz and 123 (not containing 'Y', notice XPATH is case sensitive, including xyz)}

_{5) Elements that (not) start or/and end with:}

=FILTERXML(<XML>,"//s[starts-with(., 'XY')]")

_{Returns: XY-1A, XY-3F and XY-4f (starting with 'XY')}

=FILTERXML(<XML>,"//s[not(starts-with(., 'XY'))]")

_{Returns: ABC, 123, DEF, 456, ZY-2F, xyz and 123 (don't start with 'XY')}

=FILTERXML(<XML>,"//s[substring(., string-length(.) - string-length('F') +1) = 'F']")

_{Returns: DEF, ZY-2F and XY-3F (end with 'F', notice XPATH 1.0 does not support ends-with)}

=FILTERXML(<XML>,"//s[not(substring(., string-length(.) - string-length('F') +1) = 'F')]")

_{Returns: ABC, 123, 456, XY-1A, XY-4f, xyz and 123 (don't end with 'F')}

=FILTERXML(<XML>,"//s[starts-with(., 'X') and substring(., string-length(.) - string-length('A') +1) = 'A']")

_{Returns: XY-1A (start with 'X' and end with 'A')}

_{6) Elements that are upper- or lowercase:}

=FILTERXML(<XML>,"//s[translate(.,'abcdefghijklmnopqrstuvwxyz','ABCDEFGHIJKLMNOPQRSTUVWXYZ')=.]")

_{Returns: ABC, 123, DEF, 456, XY-1A, ZY-2F, XY-3F and 123 (uppercase nodes)}

=FILTERXML(<XML>,"//s[translate(.,'ABCDEFGHIJKLMNOPQRSTUVWXYZ','abcdefghijklmnopqrstuvwxyz')=.]")

_{Returns: 123, 456, xyz and 123 (lowercase nodes)}

_{NOTE: Unfortunately XPATH 1.0 does not support upper-case() nor lower-case() so the above is a workaround. Add special characters if need be.}

_{7) Elements that (not) contain any number:}

=FILTERXML(<XML>,"//s[translate(.,'1234567890','')!=.]")

_{Returns: 123, 456, XY-1A, ZY-2F, XY-3F, XY-4f and 123 (contain any digit)}

=FILTERXML(<XML>,"//s[translate(.,'1234567890','')=.]")

_{Returns: ABC, DEF and xyz (don't contain any digit)}

=FILTERXML(<XML>,"//s[translate(.,'1234567890','')!=. and .*0!=0]")

_{Returns: XY-1A, ZY-2F, XY-3F and XY-4f (holding digits but not a a number on it's own)}

_{8) Unique elements or duplicates:}

=FILTERXML(<XML>,"//s[preceding::*=.]")

_{Returns: 123 (duplicate nodes)}

=FILTERXML(<XML>,"//s[not(preceding::*=.)]")

_{Returns: ABC, 123, DEF, 456, XY-1A, ZY-2F, XY-3F, XY-4f and xyz (unique nodes)}

=FILTERXML(<XML>,"//s[not(following::*=. or preceding::*=.)]")

_{Returns: ABC, DEF, 456, XY-1A, ZY-2F, XY-3F and XY-4f (nodes that have no similar sibling)}

_{9) Elements of certain length:}

=FILTERXML(<XML>,"//s[string-length()=5]")

_{Returns: XY-1A, ZY-2F, XY-3F and XY-4f (5 characters long)}

=FILTERXML(<XML>,"//s[string-length()<4]")

_{Returns: ABC, 123, DEF, 456, xyz and 123 (shorter than 4 characters)}

_{10) Elements based on preceding/following:}

=FILTERXML(<XML>,"//s[preceding::*[1]='456']")

_{Returns: XY-1A (previous node equals '456')}

=FILTERXML(<XML>,"//s[starts-with(preceding::*[1],'XY')]")

_{Returns: ZY-2F, XY-4f, and xyz (previous node starts with 'XY')}

=FILTERXML(<XML>,"//s[following::*[1]='123']")

_{Returns: ABC, and xyz (following node equals '123')}

=FILTERXML(<XML>,"//s[contains(following::*[1],'1')]")

_{Returns: ABC, 456, and xyz (following node contains '1')}

=FILTERXML(<XML>,"//s[preceding::*='ABC' and following::*='XY-3F']")

_{Returns: 123, DEF, 456, XY-1A and ZY-2F (everything between 'ABC' and 'XY-3f')}

_{11) Elements based on sub-strings:}

=FILTERXML(<XML>,"//s[substring-after(., '-') = '3F']")

_{Returns: XY-3F (nodes ending with '3F' after hyphen)}

=FILTERXML(<XML>,"//s[contains(substring-after(., '-') , 'F')]")

_{Returns: ZY-2F and XY-3F (nodes containing 'F' after hyphen)}

=FILTERXML(<XML>,"//s[substring-before(., '-') = 'ZY']")

_{Returns: ZY-2F (nodes starting with 'ZY' before hyphen)}

=FILTERXML(<XML>,"//s[contains(substring-before(., '-'), 'Y')]")

_{Returns: XY-1A, ZY-2F, XY-3F and XY-4f (nodes containing 'Y' before hyphen)}

_{12) Elements based on concatenation:}

=FILTERXML(<XML>,"//s[concat(., '|', following::*[1])='ZY-2F|XY-3F']")

_{Returns: ZY-2F (nodes when concatenated with '|' and following sibling equals 'ZY-2F|XY-3F')}

=FILTERXML(<XML>,"//s[contains(concat(., preceding::*[2]), 'FA')]")

_{Returns: DEF (nodes when concatenated with sibling two indices to the left contains 'FA')}

_{13) Empty vs. Non-empty:}

=FILTERXML(<XML>,"//s[count(node())>0]")

Or:

=FILTERXML(<XML>,"//s[node()]")

_{Returns: ABC, 123, DEF, 456, XY-1A, ZY-2F, XY-3F, XY-4f, xyz and 123 (all nodes that are not empty)}

=FILTERXML(<XML>,"//s[count(node())=0]")

Or:

=FILTERXML(<XML>,"//s[not(node())]")

_{Returns: None (all nodes that are empty)}

Now obviously the above is a demonstration of possibilities with XPATH 1.0 functions and you can get a whole range of combinations of the above and more! I tried to cover most commonly used string functions. If you are missing any please feel free to comment.

Whereas the question is quite broad on itself, I was hoping to give some general direction on how to use FILTERXML for the queries at hand. The formula returns an array of nodes to be used in any other way. A lot of the times I would use it in TEXTJOIN() or INDEX(). But I guess other options would be new DA-functions to spill results.