This message was deleted.
# hamilton-help
s
This message was deleted.
s
if it’s numeric, you could prefix with a letter, e.g.
n
? For
/
and
:
you could just drop in
_slash_
and
_colon_
? My only thought is to be able to reverse engineer it back incase that’s needed.
👍 1
j
yeah ive been toying around with something along those lines ... having
myprefix_<name>
and then some sort of escaping approach (i just used underscores to play around with, but being able to unescape or at least avoid conflicts is attractive)
👍 1
ive had a little bit of success with using
@parameterize_values
... it seems i can use the special chars in the parameterization which results in being able to ask for the output in the original form (ie with special chars)... i have a feeling its about to get clunky tho
e
So I definitely get that — I think the trick is to decouple the data naming/whatever from the workflow. Specifically, a mapping of files -> nodes (using a rule like the above to improve readability), then a reverse mapping allows you not to worry about how the naming maps perfectly, then assign it after you call out the nodes you want…
j
ive made some pretty good progress.... im passing in some lists of the names as config when building the dag and using
parameterize_sources
and
parameterize_values
inside a
resolve
to remap a generic function and happily sprinkling around the unsanitized names.... i can then ask a driver for an output by the name w/ special chars and it will happily use the generic function and assign the write inputs and outputs in the DAG haven't quite decided if its a good idea yet, but i think it works! ....
s
[chants blog post blog post 😁 ]