Simplify making reports in a Jupyter notebook

[jendrikseipp]

Original report by Florian Pommerening (Bitbucket: FlorianPommerening, GitHub: FlorianPommerening).

---

I tried using lab in a python notebook and wanted to create and view a report there. My hope was that I could create a report, pass a properties object to it, and get matplotlib object as a return value which would be rendered by the notebook. I tried to change the write method of report, but got stuck.

There is several parts to this suggestion. Lets discuss the general idea here and possibly create subissues then:

1. Allow reports to be run without an experiment (e.g., just given the properties)
2. Untangle the virtual methods to some degree (at least split creating the report from writing it)
3. Add the option to return the plot object (or table, or whatever is the finished version of the report) instead of writing it to a file.

[jendrikseipp]

I think it would be interesting to solve the more general problem of simplifying making reports in Jupyter notebooks. I changed the title to reflect this.

[jendrikseipp]

Original changes by Jendrik Seipp (Bitbucket: jendrikseipp, GitHub: jendrikseipp).

---

changed content from "I tried using lab in a python notebook and wanted to create and view a report there. My hope was that I could create a report, pass a properties object to it, and get matplotlib object as a return value which would be rendered by the notebook. I tried to change the write method of report, but got stuck.

There is several parts to this suggestion. Lets discuss the general idea here and possibly create subissues then:

1. Allow reports to be run without an experiment (e.g., just given the properties)
2. Untangle the virtual methods to some degree (at least split creating the report from writing it)
3. Add the option to return the plot object (or table, or whatever is the finished version of the report) instead of writing it to a file. " to "I tried using lab in a python notebook and wanted to create and view a report there. My hope was that I could create a report, pass a properties object to it, and get matplotlib object as a return value which would be rendered by the notebook. I tried to change the write method of report, but got stuck.

There is several parts to this suggestion. Lets discuss the general idea here and possibly create subissues then:

1. Allow reports to be run without an experiment (e.g., just given the properties)
2. Untangle the virtual methods to some degree (at least split creating the report from writing it)
3. Add the option to return the plot object (or table, or whatever is the finished version of the report) instead of writing it to a file."; changed title from "Create Matplotlib object given a report and properties" to "Simplify making reports in a Jupyter notebook"

[jendrikseipp]

Original comment by Florian Pommerening (Bitbucket: FlorianPommerening, GitHub: FlorianPommerening).

---

[jendrikseipp]

Original changes by Florian Pommerening (Bitbucket: FlorianPommerening, GitHub: FlorianPommerening).

---

set attachment to "jupyter.tar.xz"

[jendrikseipp]

Original comment by Florian Pommerening (Bitbucket: FlorianPommerening, GitHub: FlorianPommerening).

---

I attached an old test case of mine. I think it contains more files than are actually needed. The files Pipfile, evaluate.ipynb, and properties.xz should be enough but I cannot test that at the moment.

[FlorianPommerening]

I finally recovered the file from the backup: jupyter.zip

To try this out, run this in the directory containing the Pipfile.

pipenv install
pipenv run jupyter notebook