python calculate character length

Reprinted: http://likang.me/blog/2012/04/13/calculate-character-width-in-python/

Recently, I am writing a small CLI program in python, which involves calculating the width of characters. The goal is to truncate a long string into equal-width segments in a friendly way.

For unicode characters, python's len function can accurately calculate the number of characters contained in it, but the number does not represent the width, such as:

>>>len(u'你好a')
3

Therefore, it is not possible to simply use this method to calculate the width.

GBK decode

First of all, I thought of GBK encoding. The characters in the range of 00–7F are one-byte encoding, and the rest are two-byte encoding, which is roughly the same as the width of the characters, so there is such an opportunistic method (assuming 8 widths):

>>> a = u'hello你好'
>>> b=a.encode('gbk')
>>> try:
...   print b[:8].decode('gbk')
... except:
...   print b[:7].decode('gbk')
... 
hello你

As shown in the code, first encode the unicode string with GBK, and then try to decode it with GBK after intercepting the width of 8 bytes.

Although the problem has been initially solved, the flaws in doing so are obvious. First, the code is not elegant and runs in a trial-and-error manner; secondly, the characters that GBK can represent are limited and cannot support a large number of characters other than GBK encoding.

East_Asian_Width

After wandering around for a long time, I stumbled across the East_Asian_Width property in the Unicode Character Database standard, with the following possible values:

# East_Asian_Width (ea)

ea ; A         ; Ambiguous    不确定
ea ; F         ; Fullwidth    全宽
ea ; H         ; Halfwidth    半宽
ea ; N         ; Neutral      中性
ea ; Na        ; Narrow       窄
ea ; W         ; Wide         宽

Among them, except that A is uncertain, F/H/N/Na/W can clearly know the width. If A is regarded as a width of 2 to be conservative, it is easy to give the width of a single character:

>>> import unicodedata
>>> def chr_width(c):
...   if (unicodedata.east_asian_width(c) in ('F','W','A')):
...     return 2
...   else:
...     return 1
>>> chr_width(u'你')
2
>>> chr_width(u'a')
1

It seems that it can meet the requirements so far, but it is not uncommon to find characters with attribute A in actual use. The most typical one is Chinese double quotation marks:

>>> chr_width(u'”')
2

In most monospaced fonts, Chinese double quotation marks occupy only one digit wide. If there are multiple Chinese double quotation marks in a line, the accumulated misjudgment width will greatly reduce the interception effect, which is undoubtedly not the best. way.

urwid's solution

urwid是一个成熟的python终端UI库，它在curses的基础之上包装了类似HTML的控件用以显示文本内容，如果有这方面的开发需求，非常推荐此库，比直接使用curses库方便很多，非常棒的是它对unicode的文本宽度截取非常准确，让我大为惊讶，于是翻开它的源码一探究竟，文本宽度计算方面其核心代码如下：

widths = [
    (126,    1), (159,    0), (687,     1), (710,   0), (711,   1), 
    (727,    0), (733,    1), (879,     0), (1154,  1), (1161,  0), 
    (4347,   1), (4447,   2), (7467,    1), (7521,  0), (8369,  1), 
    (8426,   0), (9000,   1), (9002,    2), (11021, 1), (12350, 2), 
    (12351,  1), (12438,  2), (12442,   0), (19893, 2), (19967, 1),
    (55203,  2), (63743,  1), (64106,   2), (65039, 1), (65059, 0),
    (65131,  2), (65279,  1), (65376,   2), (65500, 1), (65510, 2),
    (120831, 1), (262141, 2), (1114109, 1),
]

def get_width( o ):
    """Return the screen column width for unicode ordinal o."""
    global widths
    if o == 0xe or o == 0xf:
        return 0
    for num, wid in widths:
        if o <= num:
            return wid
    return 1

如代码所示，首先根据unicode的官方EastAsianWidth文档整理出字符宽度的范围表，然后使用unicode代码查表。使用之前的例子测试：

>>> get_width(ord(u'a'))
1
>>> get_width(ord(u'你'))
2
>>> get_width(ord(u'”'))
1

完全准确，而且在实际应用中的表现也比较好，是一个理想的解决方案，更多技巧请查阅urwid的old_str_util.py源码。