Request for Feedback: Data Loading Desiderata

33 views
Skip to first unread message

Eriq Augustine

unread,
May 2, 2018, 1:56:40 PM5/2/18
to PSL Users
Hey PSL Users,

I was wondering if anyone here had any specific things they would like to see implemented with regards to data loading.
Of course the base insert methods can be used to insert any data, but that may be a bit low-level for all people.

Are there any specific formats/procedures that you would like see implemented?

-eriq

Ian Perera

unread,
May 15, 2018, 2:18:41 PM5/15/18
to PSL Users
It would be great if we could do an import of a CSV file that would have one value be constant (say, the first column), then specify additional binary predicates in each column that will take that constant value as the first argument and have the element in the column be the truth value or a second predicate.

For example, loading a Truth CSV could be like:

Day, Weekend, Weekday, Holiday
0,1,0,0
1,1,0,0
2,0,1,0
3,0,1,1

would yield predicates Weekend(0)=1, Weekend(1)=1, Weekday(2)=1,Weekday(3)=1, and Holiday(3)=1.

Something similar could also be helpful for loading in adjacency matrix forms for graphs.

Ian

bill.an...@gmail.com

unread,
Sep 17, 2018, 3:28:17 PM9/17/18
to PSL Users
Sorry for the very late reply, but one idea would be to support a prolog-like syntax that could accommodate heterogeneous predicates with optional TVs, e.g.:

person(bill) 1.0
shoe_size(bill, 12) 1.0
likes(bill, sumo) 0.9

I realize this is just sugar over the existing data loading methods (just need to bin the above representation into separate predicates and retain the arguments and TV) but it would be convenient for creating examples and in many applications where data is naturally expressed in this way.

I don't know if the existing data loaders do this, but ignoring blank lines and supporting comment syntax would be a plus in the case where the data file is intended to be read by people.

.bill

Reply all
Reply to author
Forward
0 new messages