The fact that currently, hidden columns are stored separately for each transformation is rather annoying, especially if the transformation takes long to calculate.
Test case: BNC, query string more * with Lemma_POS and Class as output columns, and a function EXTRACT(Class, "."), i.e. "S" (spoken) or "W" (written). This yields a rather large results table. If you want to get a G-test matrix for Lemma by mode, you have to switch to that transformation. But as columns can't be hidden before the data is transformed, the G-test matrix will be first calculated for all combinations of Lemma_POS1, Lemma_POS2, Class, and the function, which really takes too long.
A solution would be to treat the lsit of hidden columns as a session variable. Variables that are not available in the current transformation are ignored, while those that are available are treated as hidden.
The fact that currently, hidden columns are stored separately for each transformation is rather annoying, especially if the transformation takes long to calculate.
Test case: BNC, query string
more *
withLemma_POS
andClass
as output columns, and a functionEXTRACT(Class, ".")
, i.e. "S" (spoken) or "W" (written). This yields a rather large results table. If you want to get a G-test matrix for Lemma by mode, you have to switch to that transformation. But as columns can't be hidden before the data is transformed, the G-test matrix will be first calculated for all combinations ofLemma_POS1
,Lemma_POS2
,Class
, and the function, which really takes too long.A solution would be to treat the lsit of hidden columns as a session variable. Variables that are not available in the current transformation are ignored, while those that are available are treated as hidden.