pandas：从数字字符串中提取或拆分char

Question

I have a dataframe selected from a sql table that looks like this

   id shares_float
0   1      621.76M
1   2      329.51M

in other word,

[(1, '621.76M'), (2, '329.51M')]

I want to split the shares_float so that if it is 'B', multiply 1,000,000,000 and if it is 'M', multiply 1,000,000 and if it is neither or don't have trailing character just convert and assign the number.

Outcome should be a float type

   ticker_id  shares_float     float_value
0          1       621.76M    621760000.00
1          2         3.51B   3510000000.00

I am new to pandas. Is there a way to do it in pandas? or should I convert data to list and do my manipulation in a loop and then convert it back to pandas DataFrame?

Note added: The answer works great! Thank you. BTW, how does the function work?

Answer 1

Could use a conversion dictionary, also I am sure you didn't mean 624540000 :

In [9]:

D={'M':'*1e6', 'B':'*1e9'}
df['float_value']=df.shares_float.apply(lambda x: eval(x[:-1]+D[x[-1]]))
In [10]:

print df
   ticker_id shares_float  float_value
0          1      621.76M   621760000
1          2        3.51B  3510000000

[2 rows x 3 columns]
In [11]:

df.dtypes
Out[11]:
ticker_id         int64
shares_float     object
float_value     float64
dtype: object

Answer 2

you can use string methods to extract the pattern; for example, in order to cover all cases, starting with:

>>> df
   id shares_float
0   1            5
1   2           6M
2   3           7B

[3 rows x 2 columns]

the numeric value and unit can be extracted by:

>>> sh = df.shares_float.str.extract(r'(?P<val>[0-9.]*)(?P<unit>[MB]{0,1})')
>>> sh
  val unit
0   5
1   6    M
2   7    B

[3 rows x 2 columns]

and then:

>>> unit_map = {'':1, 'M':1e6, 'B':1e9}
>>> df['float_value'] = sh.val.astype(np.float64) * sh.unit.map(unit_map)
>>> df
   id shares_float  float_value
0   1            5            5
1   2           6M      6000000
2   3           7B   7000000000

[3 rows x 3 columns]