Modify resclust outputs

[VGPReys]

To solve the two improvments addressed in issue #327:

1. Dict keys have been casted to string.
   Tests data have been modified too.
   While printed in log, dict is converted to string, and default ' character are replaced to ", making sure the json library will be able to loads(string) the data.

2. While providing the --output=path/to/results argument, the user can now add a path where to generate the Clutering results.
   An additional test function was added to make sure the output is indeeed created while providing the argument.

[mgiulini]

nice, it works! maybe you can put also a non-json table in the output folder?
something like the standard arctic3d clustered_residues.out file

Cluster 1 -> 85 87 88 161 162
Cluster 5 -> 55 57 72 74

[rvhonorato]

@VGPReys please don't write titles like: "Solving issue #327" - instead describe what this pull request will add to the codebase 😉 and please don't make one pull request that adds more than one thing

[rvhonorato]

This will change the type of the key from integer to string and can have breaking consequences, why not use integers?

[VGPReys]

This is the point of the PR, integer as keys cannot be loaded using the json library.

[rvhonorato]

Yes this is the default behaviour of json.load, casting the type into it needs you to reclass a Decoder

[rvhonorato]

I don't agree with this and the code looks very messy.

json.loads cannot read keys as integers because this is the default json format.

This PR adds an anti-pattern solution as convenience and I don't think this is a good idea, what needs to be done is to re-think the output data structure and adopt a standard that is compliant with the default format.

Arguably changing 1 to "1" does achieve that but it relies heavy on python dynamic type casting and less on the output data formalism, which can be very prone to errors in the future.

up to you @mgiulini

[mgiulini]

I don't agree with this and the code looks very messy.

json.loads cannot read keys as integers because this is the default json format.

This PR adds an anti-pattern solution as convenience and I don't think this is a good idea, what needs to be done is to re-think the output data structure and adopt a standard that is compliant with the default format.

Arguably changing 1 to "1" does achieve that but it relies heavy on python dynamic type casting and less on the output data formalism, which can be very prone to errors in the future.

up to you @mgiulini

Hi there, sorry for the late reply.

For what I see, adding the output option that generates a clustered_residues.out file like the following

Cluster 1 -> 85 87 88 161 162
Cluster 5 -> 55 57 72 74

can be useful and is in line with the other output files. As for the json, I agree that that modification is not in the purpose of the project and of this CLI in particular. Maybe @VGPReys you can load the cluster data from the clustered_residues.out file via pandas, what do you think?

[VGPReys]

I don't agree with this and the code looks very messy.
json.loads cannot read keys as integers because this is the default json format.
This PR adds an anti-pattern solution as convenience and I don't think this is a good idea, what needs to be done is to re-think the output data structure and adopt a standard that is compliant with the default format.
Arguably changing 1 to "1" does achieve that but it relies heavy on python dynamic type casting and less on the output data formalism, which can be very prone to errors in the future.
up to you @mgiulini

Hi there, sorry for the late reply.

For what I see, adding the output option that generates a clustered_residues.out file like the following

Cluster 1 -> 85 87 88 161 162
Cluster 5 -> 55 57 72 74

can be useful and is in line with the other output files. As for the json, I agree that that modification is not in the purpose of the project and of this CLI in particular. Maybe @VGPReys you can load the cluster data from the clustered_residues.out file via pandas, what do you think?

Ok then let's just stick to the textual version in clustered_residues.out.