Data.name ends up as the actual file name in the actual command line arguments to a ShellTask

[GeigerJ2]

NOTE: Currently working off #126.

I'd assume the actual name in the YAML file is just a placeholder to link inputs/outputs of the cycles section to the data section? How should the actual file/directory name be determined? I'd say from the src provided in the relevant data entry.

E.g., using initial_cond in the YAML file leads to the following calcinfo.json (this file is dumped by AiiDA in the working directory of the process in the hidden .aiida folder, and contains instructions on the execution of the process):

{
  "codes_info": [
    {
      "code_uuid": "5f4a3f0d-160e-4f1d-b778-e8d584ce1b0f",
      "cmdline_params": ["--restart", "--init", "initial_cond"],
      "stdin_name": null,
      "stdout_name": "stdout",
      "stderr_name": "stderr"
    }
  ],
  "append_text": "echo $? > status",
  "remote_copy_list": [],
  "remote_symlink_list": [
    [
      "da8379b9-0b77-44ad-9dc2-6b9a03d7ca2b",
      "/mnt/home/geiger_j/test/sirocco/data/initial_conditions/*",
      "."
    ]
  ],
  "retrieve_temporary_list": [
    "restart",
    "icon_output",
    "status",
    "stderr",
    "stdout"
  ],
  "provenance_exclude_list": ["input"],
  "file_copy_operation_order": [1, 0, 2],
  "uuid": "7683f368-c4e5-4be0-964e-28d58b88dc60"
}

This could be fixed by setting the filenames argument to the ShellJob/ShellTask.
Alternatively, fix could be applied in the resolve_ports method here:
https://github.com/GeigerJ2/Sirocco/blob/ca2152ec2f72c349947db22db0b14bf52280007c/src/sirocco/parsing/yaml_data_models.py#L286
-> Probably not, though, as this method still yields placeholders...

[GeigerJ2]

As discussed with @agoscinski, there are 3 options:

1. We first make a quick fix in aiida-shell to check if RemoteData points to file or dir and modify the instructions for the remote_[copy,symlink]_list accordingly.
2. We extend the API of RemoteData (at least add is_dir/is_file). However, as we take much longer with releases in aiida-core a fix in aiida-shell is probably simpler for now.
3. In addition, even though the instruction in the remote_symlink_list doesn't make any sense, we do not get an error in aiida-core. The problem here is that we construct the RemoteData from the absolute path to a file, while in aiida-shell, when the instruction of the CalcInfo is created in the prepare_for_submission, this path is expanded by a tailing /*.

Other, hacky possible solution in Sirocco: Use filenames argument, either resolved automatically from src, or expose to the user, but preferred that we fix this in AiiDA.

[GeigerJ2]

Should be fixed by sphuber/aiida-shell#109.