使这个数据清理循环更多Python

Question

I am cleaning some human classified data with Python 2.7, mainly using pandas, but using numpy.isreal() to check for floats because some people apparently entered floats in fields like 'background_color' . Anyway, I'm posting an example of what it would look like for one color with my current set up, which works, it just doesn't look very Python, by the end of the loops, blues is a list of all the indexes where background_color was case insensitive 'BLUE' :

blueShapes=[]
for i in range(imageData.shape[0]):
    if not (np.isreal(imageData.loc[i,'background_color'])):
        if imageData.loc[i,'background_color'].upper()=='BLUE':
            blueShapes.append(i)

It seems like I could use the map function to make this more Pythonic and prettier. Like I said, it functions as intended, but just seems too...C or Java for it to be written in Python. Thanks in advance for any responses.

-Edit: I removed the count because it was a relic from an old loop

Answer 1

you could create a bew column with upper cases

imageData['background_color_2'] = map(lambda x: x.upper(), imageData['background_color'].astype(str))

subset = imageData[imageData['background_color_2']=='BLUE']

for the count

len(subset['background_color'])

Answer 2

You can define a lambda function that returns the index of rows with a specific string value

getRowIndexWithStringColor = lambda df, color: [i for i in range(df.shape[0]) if (not np.isreal(df.loc[i,'background_color'])) and df.loc[i,'background_color'].upper()==color)]
rowIndexWithBlue = getRowIndexWithStringColor(imageData, 'BLUE')

Answer 3

As a general rule if you are looping in pandas you are doing it wrong.

Should look something like this (though untested so you need to adapt it!):

strings = (~imageData.background_color.apply(np.isreal))
blue = (imageData.background_color.str.upper()=="BLUE")
blueshapes = imageData[strings & blue].index           

Answer 4

Thanks all! I used a minor adaptation to Steven G's answer, I have all of this backed up in a master .csv, so I had no quams with just overwriting the background_color column with its string equivalent. Any non-sring entries are invalid anyway, but they are not alone so I will find them later as leftover indexes after i concatenate the indexes of all the colors. Each list will be extracted as follows:

imageData['background_color']=map(lambda x: x.upper(), imageData['background_color'].astype(str))

blueShapes=imageData[imageData['background_color']=='BLUE'].index

Answer 5

I would turn it into a function and return an array.

Google: Zen of Python

Giving you a quick reference python list/dict/set over map/filter .

Better readability and cleaner code.

def colorShapes(color):
    return [i
             for i in range(imageData.shape[0])
             if not(np.isreal(imageData.loc[i, 'background_color'].upper() == color 
             and imageData.loc[i, 'background_color'].upper() == color]