How to Create a Inputs File
An inputs json file has the below format.
Typically, you pass a variable to the workflow and within the WDL, you can pass the variable to one or more tasks.
{
"<workflow name>.<variable name>": "<value>"
}
Alternatively, you can pass a variable to an individual task
{
"<workflow name>.<task name>.<variable name>": "<value>"
}
You can create an inputs file by scratch or you can build a template based on the WDL using the following command:
jaws inputs <path to your.wdl>
This command should output a template for the inputs.json file. You can then fill in the values of each key.
{
"bbtools.reads": "File",
"bbtools.ref": "File"
}
Note
File paths may be absolute or relative paths. Any string in the inputs json file is accepted as a file as long as it has been declared as a File type in the corresponding WDL. File paths that start with http and ftp are recognized as URLs.
To include lists or dictionaries in your input.json files, you would do something like:
{
"bbtools.var_list": ["a","b","c"]
"bbtools.var_map": { "first": "a",
"second": "b",
"third": "c" }
}
Note
JAWS will cache your input files for some number of days (e.g. 14d) so that if you use the same file in another run (e.g. reference data file), it will not need to be transferred. Your input files are not deleted so long as they are accessed within this time period.
Note
Some compute nodes have fast local disk (e.g. SSD) or large local disk (e.g. HD). If available, they will be mounted as /fast_scratch and/or /big_scratch, respectively. The availability varies between compute sites.