Bonjour Giuseppe, My first guess is that count($p) is a very very huge number ? In the first query you will scan the $p sequence at each iteration. In the second one you will directly access the text index. Hoping it helps Best regards, Fabrice -----Message d'origine----- De : basex-talk-bounces@mailman.uni-konstanz.de [mailto:basex-talk-bounces@mailman.uni-konstanz.de] De la part de Giuseppe Celano Envoyé : mardi 19 septembre 2017 15:35 À : BaseX Objet : [basex-talk] db:text() vs XPath I am using BaseX 8.6.4 and I am trying to do a group-by/order-by operation, and I see that two logically equivalent queries perform very differently: one cannot see the end, while the other can (and fast). I can provide further details if necessary, but these are the queries (look at the last line of both, where the difference is): (This does not work:) declare variable $p := db:open("db2")/text/line/text(); (: returns a list of values like ΠΕΡΙ#n-s---mv- :) for $u in db:open("db1")/text/s/t let $a := $u/f/text() (: returns one value like ΠΕΡΙ :) let $b := $u/@o/data(.) (: returns one value like n-s---mv- :) group by $k := $a || "#" || $b (: builds a value like ΠΕΡΙ#n-s---mv- :) let $n := count($u) order by $n descending return if ($p = $k) then $k || " " || $n else () (This works:) for $u in db:open("db1")/text/s/t let $a := $u/f/text() (: returns one value like ΠΕΡΙ :) let $b := $u/@o/data(.) (: returns one value like n-s---mv- :) group by $k := $a || "#" || $b (: builds a value like ΠΕΡΙ#n-s---mv- :) let $n := count($u) order by $n descending return if (db:text("db2", $k)) then $k || " " || $n else () The problem is with directly accessing the database with db:text() or indirectly using XPATH. I would tend to use XPATH and expect to get an underlying translation into db:text(), but this seems not to happen: why? Best, Giuseppe

Hi Giuseppe, The following query may be faster: for $u in db:open("db1")/text/s/t group by $k := $u/f || "#" || $u/@o where db:open("db2")/text/line[text() = $k] let $n := count($u) order by $n descending return $k || " " || $n Cheers, Christian On Tue, Sep 19, 2017 at 3:35 PM, Giuseppe Celano <celano@informatik.uni-leipzig.de> wrote:

I am using BaseX 8.6.4 and I am trying to do a group-by/order-by operation, and I see that two logically equivalent queries perform very differently: one cannot see the end, while the other can (and fast). I can provide further details if necessary, but these are the queries (look at the last line of both, where the difference is):

(This does not work:)

declare variable $p := db:open("db2")/text/line/text(); (: returns a list of values like ΠΕΡΙ#n-s---mv- :)

for $u in db:open("db1")/text/s/t let $a := $u/f/text() (: returns one value like ΠΕΡΙ :) let $b := $u/@o/data(.) (: returns one value like n-s---mv- :) group by $k := $a || "#" || $b (: builds a value like ΠΕΡΙ#n-s---mv- :) let $n := count($u) order by $n descending return if ($p = $k) then $k || " " || $n else ()

(This works:)

for $u in db:open("db1")/text/s/t let $a := $u/f/text() (: returns one value like ΠΕΡΙ :) let $b := $u/@o/data(.) (: returns one value like n-s---mv- :) group by $k := $a || "#" || $b (: builds a value like ΠΕΡΙ#n-s---mv- :) let $n := count($u) order by $n descending return if (db:text("db2", $k)) then $k || " " || $n else ()

The problem is with directly accessing the database with db:text() or indirectly using XPATH. I would tend to use XPATH and expect to get an underlying translation into db:text(), but this seems not to happen: why?

Best, Giuseppe