Open potash opened 8 years ago
@potash apologies, without an example workflow i'm not sure what might be happening here. the best i can do is point you to some historic literature that may be related: https://github.com/Factual/drake/pull/149
Here's an example workflow that recreates the problem:
c <- z, b
cp $INPUT $OUTPUT
a <- x
cp $INPUT $OUTPUT
b <- y, a
cp $INPUT $OUTPUT
First running touch x y z
, and then drake
, I get:
$ drake
java.lang.Exception: no input data found in locations: [...]/./a
However, this workflow runs without a problem if I call drake ...
.
Alternatively, I can fix this workflow by changing the order of the steps and just calling drake
(without ...
); this workflow works then:
a <- x
cp $INPUT $OUTPUT
b <- y, a
cp $INPUT $OUTPUT
c <- z, b
cp $INPUT $OUTPUT
The manual states
I'm confused-- is
...
the default or not?Apparently it is not because I a workflow such that if I run
drake
I get some "input not found" errors (for inputs which are outputs of steps in the workflow!) but the error goes away when I rundrake ...
.Unfortunately the workflow is complicated and I have been unsuccessful in creating a simple shareable example. Any thoughts on why
drake
might fail to find an output whiledrake ...
works? Thanks!