sddk 1.1

A package to access sciencedata.dk

sddk

This is a simple Python package to write files to- and read files from sciencedata.dk. It is especially designed for working with shared folders. It relies mainly upon Python requests library.

sciencedata.dk is a project managed by DEiC (Danish e-infrastrcture cooperation) aimed to offer a robust data storage, data management and data publication solution for researchers in Denmark and abroad (see docs and dev for more info). The storage is accessible either through (1) the web interface, (2) WebDAV clients or (3) an API relaying on HTTP Protocol (see docs and dev for more info). One of the strength of sciencedata.dk is that it currently supports institutional login from 2976 research and educational institutions around the globe (using WAYF). That makes it a perfect tool for international research collaboration.

The main functionality of the package is in uploading any Python object (dict, list, dataframe) as a text or json file to a preselected shared folder and getting it back into a Python environemnt as the original Python object. It uses sciencedata.dk API in combination with Python requests library.

Install and import

To install and import the package within your Python environment (i.e. jupyter notebook) run:

!pip install sddk ### to be updated, use flag --ignore-installed
from sddk import * ### import all functions

Session configuration

To run the main configuration function below, you have to know the following:

• your sciencedata.dk username (e.g. "[e-mail]" or "[e-mail]"),
• your sciencedata.dk password (has to be previously configured in the sciencedata.dk web interface),

In the case you want to access a shared folder, you further need:

• name of the shared folder you want to access (e.g. "our_shared_folder"),

• username of the owner of the folder (if it is not yours)

(Do not worry, you will be asked to input these values interactively while running the function)

To configure a personal session, run:

s, sddk_url = sddk.configure_session_and_url()

Configuration of a session with shared folder

To configure a session pointing to a shared folder, run:

s, sddk_url = sddk.configure_session_and_url("our_shared_folder", "[e-mail]")

Running this function, you configurate two key variables:

• s: a request session authorized by your username and password
• sddk_url: default url address (endpoint) for your request Below you can inspect how these two are used in typical request commands

write_file()

The most important components of the package are two continuously developed functions: write_file(path_and_filename, python_object) and read_file(path_and_filename, type_of_object).

So far the function has been used with several different types of Python objects: str, list, dictionary, Pandas' dataframe and Matplotlib's figure. These can been written either as .txt, .json or .png files, based simply upon the filename's ending chosen by the user. Here are simple instances of these python objects to play with:

### Python "str" object
string_object =  "string content"
### Python "list" object
list_object = ['a', 'b', 'c', 'd']
### Python "dictionary" object
dict_object = {"a" : 1, "b" : 2, "c":3 }
### Pandas dataframe object
import pandas as pd
dataframe_object = pd.DataFrame([("a1", "b1", "c1"), ("a2", "b2", "c2")], columns=["a", "b", "c"]) 
### Matplotlib figure object
import matplotlib.pyplot as plt
figure_object = plt.figure() # generate object
plt.plot(range(10)) # fill it by plotted values

The simplest example is once we want to write a string object into a textfile located at our home folder (Remember, that since the configuration this home folder is contained within the sddk_url variable )

write_file("test_string.txt", string_object)

In the case that everything is fine, you will receive following message:

> Your <class 'str'> object has been succefully written as "https://sciencedata.dk/files/test_string.txt"

However, there is a couple of things which might go wrong. You can choose an unsupported python object, a non-existent path or unsupported file format. The function captures some of these cases. For instance, once you run write_file("nonexistent_folder/filename.wtf", string_object), you will be interactively asked for corrections. First: the function checks whether the path is correct. When corrected to an existent folder (here it is "personal_folder"), the function further inspect whether it has known ending (i.e. txt, json or png). If not, it asks you interactively for correction. Third, it checks whether the folder already contain a file of the same name (to avoid unintended overwritting), and if yes, asks you what to do. Finally, it prints out where you can find your file and what type of object it encapsulates.

>>> The path is not valid. Try different path and filename: personal_folder/textfile.wtf
>>> Unsupported file format. Type either "txt", "json", or "png": txt
>>> A file with the same name ("textfile.txt") already exists in this location.
Press Enter to overwrite it or choose different path and filename: personal_folder/textfile2.txt
>>> Your <class 'str'> object has been succefully written as "https://sciencedata.dk/files/personal_folder/textfile2.txt"

The same function works with dictionaries, lists, Matplotlib's figures and especially Pandas' dataframes. Pandas' dataframe is my favorite. I send there and back 1GB+ dataframes as json files on a daily basis.

read_file()

On the other side, we have the function read_file(path_and_filename, object_type), which enables us to to read our files back to python as chosen python objects. Currently, the function can read only textfiles as strings, and json files as either dictionary, lists or Pandas's dataframes. You have to specify the type of object as the second argument, the values are either "str", "list", "dict" or "df" within quotation marks, like in these examples:

string_object = read_file("test_string.txt", "str")
string_object
>>> 'string content'

list_object = read_file("simple_list.json", "list")
list_object
>>> ['a', 'b', 'c', 'd']

dict_object = read_file("simple_dict.json", "list")
dict_object
>>> {'a': 1, 'b': 2, 'c': 3}

dataframe_object = read_file("simple_df.json", "df")
>>>     a   b   c
0  a1  b1  c1
1  a2  b2  c2

PUT and GET requests in detail

In the core of thewrite_file()function is the PUT request command. Here is how what it basically does in the case of different types of objects:

String to TXT

Upload (export) simple text file:

s.put(sddk_url + "testfile.txt", data="textfile content")

Get it back (import) to Python:

string_testfile = ast.literal_eval(s.get(sddk_url + "testfile.txt").text)
print(string_testfile)

Pandas DataFrame to JSON

Upload a dataframe as a json file:

import pandas as pd
df = pd.DataFrame([("a1", "b1", "c1"), ("a2", "b2", "c2")], columns=["a", "b", "c"]) 
s.put(sddk_url + "df.json", data=df.to_json())

Get it back:

df = pd.DataFrame(s.get(sddk_url + "df.json").json())

Pandas DataFrame to CSV

import pandas as pd
df = pd.DataFrame([("a1", "b1", "c1"), ("a2", "b2", "c2")], columns=["a", "b", "c"]) 
df.to_csv("df.csv") ### temporal file
s.put(sddk_url + "df.csv", data = open("df.csv", 'rb'))

Dictionary to JSON

To sciencedata.dk:

dict_object = {"a" : 1, "b" : 2, "c":3 }
s.put(sddk_url + "dict_file.json", data=json.dumps(dict_object))

From sciencedata.dk:

dict_object = json.loads(s.get(sddk_url + "dict_file.json").content)

Matplotlib figure to PNG

import matplotlib.pyplot as plt
fig = plt.figure()
plt.plot(range(10))
fig.savefig('temp.png', dpi=fig.dpi) ### works even in Google colab
s.put(sddk_url + "temp.png", data = open("temp.png", 'rb'))

Next steps

• to make the functions more robust.

def read_file(file_name_and_loc, type_of_python_object):
  if s.get(sciencedata_groupurl + file_name_and_loc).ok:
    print("file exists")
    try: 
      return json.loads(s.get(sciencedata_groupurl + file_name_and_loc).content) ### if there already is a file with the same name
    except:
      print("file import failed")
  else:
    print("file does not found; check file name and path.")

The package is built following this tutorial.

Versions history

• 0.0.6 - first functional configuration
• 0.0.7 - configuration of individual session by default
• 0.0.8 - shared folders reading&writing for ordinary users finally functional
• 0.1.1 - added shared folder owner argument to the main configuration function; migration from test.pypi to real pypi
• 0.1.2 - added redirection
• 1.0 - added functions write_file() and read_file()
• 1.1 - a simple correction