Skip to content

Instantly share code, notes, and snippets.

@Dan-Patterson Dan-Patterson/
Last active Nov 19, 2018

What would you like to do?
This can be used to convert excel files to numpy structured or record arrays.
# see for
# more information and descriptions
import numpy as np
def excel_np(path, sheet_num=0):
"""Read excel files to numpy structured/record arrays. Your spreadsheet
must adhere to simple rules::
- first row must contain the field names for the output array
- no blank rows or columns, basically, no fluff or formatting
- if you have nodata values, put them in, since blank cells will be
'corrected' as best as possible.
- text and numbers in a column, results in a text column
path : text
Full path to the xls, xlsx file
sheet_num : integer
Sheets are numbered from 0.
A numpy structured array is returned. Excel only uses float or string
data, so attempts are made to coerse integer columns by comparing the
float vs int versions of the arrays. A tad of a kludge, but it works.
The first row's data type is compared to its matching column data type.
If they match, then it is used as the dtype. If there is a mismatch an
attempt is made to recover numeric data by assigning blanks etc in numeric
columns a value of np.nan.
String/text columns are check for empty cells, '', "" and that ever so
ugly invisible space.
>>> aString = open('c:/temp/test.xlsx','rb').read()
>>> book_ = open_workbook(file_contents=aString)
>>> dir(book_):
get_sheet, nsheets, sheet_by_index, sheet_by_name etc....
Now you can read a sheet
>>> sheet = book_.sheet_by_index(0) # first sheet
>>> sheet.col_types(0)
def _values(sheet, rows, cols):
"""return cell types for the above. Skip the first row
Not use.... just kept for future reference
ar = []
for i in range(1, rows):
c = []
for j in range(cols):
c.append(sheet.cell_values(i, j)) # sheet.cell_types also
return ar
def isfloat(a):
"""float check"""
i = float(a)
return i
except ValueError:
return np.nan
def punc_space(name):
"""delete punctuation and spaces and replace with '_'"""
punc = list('!"#$%&\'()*+,-./:;<=>?@[\\]^`{|}~ ')
return "".join([[i, '_'][i in punc] for i in name])
import xlrd
w = xlrd.open_workbook(path) # class
sheet = w.sheet_by_index(sheet_num) # sheet by number
# sheet = w.sheet_by_name('test') # case sensitive, not implemented
names = sheet.row_values(0) # clean these up later
cols = sheet.ncols
rows = sheet.nrows
col_data = [sheet.col_values(i, 1, rows) for i in range(cols)]
row_guess = sheet.row_values(1)
row_dts = [np.asarray(i).dtype.kind for i in row_guess]
col_dts = [np.asarray(col_data[i]).dtype.kind
for i in range(cols)]
clean = []
for i in range(len(row_dts)):
c = col_data[i]
if row_dts[i] == col_dts[i]: # same dtype... send to array
ar = np.asarray(c)
if row_dts[i] == 'f': # float? if so, substitute np.nan
ar = np.array([isfloat(i) for i in c])
is_nan = np.isnan(ar) # find the nan values, then check
not_nan = ar[~is_nan] # are the floats == ints?
if np.all(np.equal(not_nan, not_nan.astype('int'))): # integer?
ar[is_nan] = -999
ar = ar.astype('int')
elif row_dts[i] in ('U', 'S'): # unicode/string... send to array
ar = np.char.strip(ar)
ar = np.where(np.char.str_len(ar) == 0, 'None', ar)
ar = np.asarray(c)
# ---- assemble the columns for the array ----
dt_str = [i.dtype.str for i in clean]
names = [i.strip() for i in names] # clean up leading/trailing spaces
names = [punc_space(i) for i in names] # replace punctuation and spaces
dts_name = list(zip(names, dt_str))
arr = np.empty((rows-1,), dtype= dts_name)
cnt = 0
for i in names:
arr[i] = clean[cnt]
cnt +=1
return arr
# ----------------------------------------------------------------------
# __main__ .... code section
if __name__ == "__main__":
: provide a spreadsheet for testing
#path = "c:/test/text.xlsx"
#arr = excel_np(path, 0)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.