Make io functions accept file like objects

[rwl]

Is your feature request related to a problem? Please describe.
Currently, if I have data in memory that I wish to parse into a System it has to be written out to a temporary file and the path passed to a read function from an io module.

Describe the solution you'd like
It would be good if the io modules had read_file functions that would accept file-like objects (e.g. StringIO). These could be called from the existing read functions that would still take paths and handle opening/closing files.

Describe alternatives you've considered
The xlsx module already supports this as the Pandas ExcelFile constructor accepts file-like objects. The json module is simple enough to just copy, but the matpower and psse modules call open.

[cuihantao]

Typed on phone, please excuse the formatting.

I will work on refactoring the parsers.

For the xlsx reader, I have to switch back to pandas.read_excel to specify openpyxl as the engine, since xlrd deprecated the support for xlsx file. But that shouldn’t make a difference for the purpose of reading from a file object.

[cuihantao]

Fixed it in dfeea93.

Added a wrapper read_file_like(infile) in andes.io.__init__ to accept a file-like or a str for path.
Modified each parser to use it instead of directly calling open.

To create a system object and load from file, use

import andes

fd = open(andes.get_case("npcc/npcc.raw"))

ss = andes.System()  # args can go here.
ss.undill()
andes.io.psse.read(ss, fd)
ss.setup()

fd.close()

In order for Andes to output correctly, you will need to call ss.files.set("presumed_case_name") manually. I hope you won't mind checking the code in andes/variables/fileman.py.

Let me know if you have other suggestions.

[rwl]

Thank you very much for making these changes so quickly. This will be much more convenient than writing out to temporary files.

[cuihantao]

You are welcome. As always, thank you for the suggestion.