improved error message

[nsheff]

If I'm a naive user and just testing things out, my first inclination, without reading docs, is to do somethign like this:

pip install pipestat
pipestat report -i myval -v 123

The error I get is this:

pipestat report -i myval -v 123
Traceback (most recent call last):
  File "/home/nsheff/.local/bin/pipestat", line 8, in <module>
    sys.exit(main())
  File "/home/nsheff/.local/lib/python3.8/site-packages/pipestat/cli.py", line 28, in main
    psm = PipestatManager(
  File "/home/nsheff/.local/lib/python3.8/site-packages/pipestat/pipestat.py", line 142, in __init__
    self[NAME_KEY] = _select_value(
  File "/home/nsheff/.local/lib/python3.8/site-packages/pipestat/pipestat.py", line 112, in _select_value
    raise PipestatError(
pipestat.exceptions.PipestatError: Value for the required 'namespace' argument could not be determined. Provide it in the config or pass to the object constructor.

I think it should instead say something like "No config file provided. you must provide a pipestat config file" or something.

[nsheff]

related, since the config is required to run pipestat, maybe should be explalined here:

http://pipestat.databio.org/en/latest/install/

[stolarczyk]

Ok, I'll need to work on messaging in the CLI. In general, the config isn't required, e.g:

In [1]: import pipestat                                                                              

In [2]: psm = pipestat.PipestatManager(namespace="test", results_file_path="/Users/mstolarczyk/Deskto
   ...: p/testing/test.yml")                                                                         
Initializing results file '/Users/mstolarczyk/Desktop/testing/test.yml'

In [3]: print(psm)                                                                                   
In [1]: import pipestat                                                                              

In [2]: psm = pipestat.PipestatManager(namespace="test", results_file_path="/Users/mstolarczyk/Deskto
   ...: p/testing/test.yml")                                                                         
Initializing results file '/Users/mstolarczyk/Desktop/testing/test.yml'

it is required if a DB is the backend because that's the only way to provide the login credentials.

[vreuter]

@nsheff can you provide more clarity about the relationship between writing to database and writing to file? I've been looking at this, and it's not yet clear if these are mutually exclusive, and which things are conditional on using the database as a backend, and even how using a database as a backend is determined. Is it exclusively by presence of a config (file or mapping)? If so, I think that should be renamed to be more indicative of the fundamental backend change that providing a value there is effecting.

[vreuter]

^^From the end of PipestatManager.__init__, particularly the if/elif/else block, it looks like the database and file as backend are indeed mutually exclusive. I'm not sure, however, how to use this in the resolution of this issue, as the relationship between calling _select_value to try to get "namespace" (as is where the error in this issue occurs) and the presence / absence of a config file is unclear.

[vreuter]

@nsheff what do you think about this?

(pipestat) vince@vr-think:~/code/pipestat$ pipestat report -i myval -v 123
Value for the required 'namespace' argument could not be determined.
Traceback (most recent call last):
  File "/home/vince/venvs_python/pipestat/bin/pipestat", line 8, in <module>
    sys.exit(main())
  File "/home/vince/venvs_python/pipestat/lib/python3.8/site-packages/pipestat/cli.py", line 29, in main
    psm = PipestatManager(
  File "/home/vince/venvs_python/pipestat/lib/python3.8/site-packages/pipestat/pipestat.py", line 155, in __init__
    raise PipestatError(msg)
pipestat.exceptions.PipestatError: Could not determine value for namespace. This may be passed directly to the pipestat manager constructor, or through a configuration file, or as an environment variable: PIPESTAT_NAMESPACE

[nsheff]

@nsheff can you provide more clarity about the relationship between writing to database and writing to file? I've been looking at this, and it's not yet clear if these are mutually exclusive, and which things are conditional on using the database as a backend, and even how using a database as a backend is determined. Is it exclusively by presence of a config (file or mapping)? If so, I think that should be renamed to be more indicative of the fundamental backend change that providing a value there is effecting.

database and file back-end are mutually exclusive. You just choose one. pipestat provides an interface that can read and write from either of them. I think the config file specifies which you are doing

[nsheff]

@nsheff what do you think about this?

pipestat.exceptions.PipestatError: Could not determine value for namespace. This may be passed directly to the pipestat manager constructor, or through a configuration file, or as an environment variable: PIPESTAT_NAMESPACE

This seems very developer-oriented.

I would think some kind of messaging that says "back-end not specified" would make more sense here

[vreuter]

I would think some kind of messaging that says "back-end not specified" would make more sense here

I don't see how the two things (one being this error, the other being the lack of specification of a back-end) are related. This error appears to happen when two things are jointly true:
a) namespace is not passed directly to the constructor of the PipestatManager
b) The namespace can be determined neither from the config file nor from the environment variable.

To me, it seems that at this point in the logic, where the error's occurring, there's not yet been any inspection of whether a results file path is available (which seems to signal use of the file as backend, rather than database); specifically, nothing at this point (where the error's occurring) in the logic of the constructor has dealt with the value of results_file_path.

What am I missing?

[vreuter]

^^ @nsheff perhaps I don't yet fully understand the different ways in which a config file could be formatted / used, but the error here seems to be just related to retrieval of a specific value from either the environment or the config file; it seems to me like a user encountering this may well have passed a value for results_file_path, thereby specifying a backend (to file).

[vreuter]

From discussion with @nsheff in infra meeting, the precipitating event here (i.e., the command-line use case presented to start the ticket) should fail because of lack of specification of a backend.