pd.io.gspread.read_frame - from Google Spreadsheet to Pandas DataFrame

Google Spreadsheet is an online spreadsheet. Google Document can be use to generate survey and results of survey will be stored as a Google Spreadsheet document.

Maybe Pandas should provide a pd.io.gspread.read_frame function that will read a given Google Spreadsheet document (using email, password, url or name of document and range) and return a DataFrame.

pd.io.gspread.read_frame(email, password, filename, sheet, cell_range)

gspread package could help http://burnash.github.io/gspread/

I wrote a little bit of code for that… but it could probably be improve and add into Pandas.

email = '...@...'
password = '...'
cell_range = 'A1:R20'

gc = gspread.login(email, password)
wks = gc.open(filename).sheet1

cell_list = wks.range('')

# Build a NumPy array
(row, col) = (cell_list[-1].row, cell_list[-1].col)
data = np.empty((row-1,col), dtype=object)
data[:] = np.nan

k = 0
cols = []
for i in range(row):
    for j in range(col):
        val = cell_list[k].value
        if i==0:
            if val != None:
                if val not in cols:
                    cols.append(val)
                else: # add a number if colname ever exists
                    ii = 1
                    while True:
                        new_val = val + '_' + str(ii)
                        if new_val not in cols:
                            break
                        ii += 1
                    cols.append(new_val)
            else:
                cols.append('col_'+str(j))
                #cols.append(j)
        else:
            if val != None:
                data[i-1, j] = val
        k += 1

[jtratner]

Interesting. I’m not totally clear on what you’re trying to do above - why are you special casing row 0 (instead of just doing something with row 0 first and then changing the line to: for i in range(1, row): instead? It’s also unclear why you’re using k and j in a for loop, but maybe I’m missing something.

Based on a quick read of gspread, the easiest thing to do would be something like:

values = wks.get_all_values()
header, rows = values[0], values[1:]
df = DataFrame(rows, columns=header)

That said, the above is particularly inefficient because it first stores all of the cells in Cell objects, creates a defaultdict, stores everything, then uses a double list comprehension to get all the data.

1 big question: Why would you use gspread over the Google Data API? Are their particular advantages?

[maybelinot]

Here is library that provides possibilities for uploading/downloading between spreadsheets and pandas DataFrame, for those who tries to find solutions in this issue.