尝试使用ftplib和storbinary ==> UnicodeEncodeError存储png图像

Question

I'm new to StackOverflow so I'm really excited! 😊

I'm having the following issue with my code: I am trying to store a 'png' image into an FTP server (a screenshot of a website).

I'm using ftplib and selenium (with webdriver):

driver.get(<someURL>)
screenshot = driver.save_screenshot(driver.title + '.png')
ftp.storbinary("STOR <PathToServer>" + driver.title + '.png', open(driver.title + '.png', 'rb'))

This method works when the website is written with Latin characters, the problem is that the image can be the screenshot of a website located in Thailand, China, or Egypt, for instance. In that case, the line with:

open(driver.title + '.png', 'rb')

returns the infamous error:

UnicodeEncodeError: 'latin-1' codec can't encode character '\u1ec7' in position 60: ordinal not in range(256)

I understand that storbinary is accepting only binary numbers (as the name of that method implies). However, what I don't understand is how I can "encode" the png so that it will not lead to that error, and so that it can be stored successfully into the FTP server.

Thank you so very much! Any help, or comment, or insight, would be deeply appreciated. Best!

Answer 1

Fact is text as represented in a running computer program and as stored in files, filenames and databases are two different things. One can think of the former as a set of characters, without minding how they are represented internally. On the other hand, to store this text in a filesystem, DB, transmit it over a network, the text have to be represented as bytes. This process of transforming the "pure" text you have when the program is running into a byte representation is called "encoding". For better understanding that, I suggest reading this article .

Python 3, both the core language and the libraries, try to automatically select the proper text encoding when doing any I/O with text. In your case, it picked the "latin1" codec for filenames in the target server.

Latin1 is limited to a little over 200 valid characters and can't represent a lot of characters or glyphs - any non-western language character,and even some western ones, such as Ĺ, Ṕ, ŵ, can't be represented with it.

The suggestion is to perform a manual encoding of the name before leting Python doing so, because then we can have control on how to handle non-existing characters in the target encoding. Since the library method ( .strobinary ) seems to be expecting the filename as a string, then, we "decode" the name back, but keeping the replacements for invalid characters we got when first encoding, and pass the result of this roundtrip to the library.

So, to keep the information about your characters that does not exist in latin1, I'd suggest using an escape-encoding - other options would be to replace then with a "?" or ignore, just supressing all characters:

filename = driver.title.encode("latin1", errors="xmlcharrefreplace").decode("latin1") + ".png"
screenshot = driver.save_screenshot(filename)
ftp.storbinary("STOR <PathToServer>" + filename, open(filename, 'rb'))