<blockquote>
<p>Why does json.dumps escape non-ascii characters with “\uxxxx”</p>
</blockquote>
<p>Python 2可以将ascii-only-bytestrings和Unicode字符串混合在一起。</p>
<p>这可能是一个过早的优化。在Python 2中,如果Unicode字符串包含的字符大多在ASCII范围内,则它可能需要比相应的bytestring多2-4倍的内存。</p>
<p>另外,即使在今天,如果打印到Windows控制台时包含非ascii字符,则<code>print(unicode_string)</code>可能很容易失败,除非安装了类似<code>win-unicode-console</code>的Python包。如果使用C/POSIX语言环境(在许多情况下,<code>init.d</code>服务的默认值,<code>ssh</code>,<code>cron</code>),即使在Unix上也可能失败(这意味着使用ascii字符编码)。有<code>C.UTF-8</code>但是它并不总是可用的,您必须显式地配置它)。这也许可以解释为什么在某些情况下您可能需要<code>ensure_ascii=True</code>。</p>
<p>JSON格式是为Unicode文本定义的,因此严格来说,<code>json.dumps()</code>应该始终返回Unicode字符串,但如果所有字符都在ASCII范围内(<code>xml.etree.ElementTree</code>具有类似的“优化”),则它可能返回bytestring。Python 2允许在某些情况下(允许隐式转换)将ascii-only-bytestring视为Unicode字符串,这令人困惑。Python 3更严格(禁止隐式转换)。</p>
<p><strong>可以使用ASCII-only bytestrings代替Unicode字符串(可能有非ASCII字符)来节省内存和/或提高Python 2中的互操作性。</strong></p>
<p>要禁用该行为,请使用<code>json.dumps(obj, ensure_ascii=False)</code>。</p>
<hr/>
<p>避免将Unicode字符串与其在Python源代码中的<em>表示形式</em>混淆为Python字符串文本或将其在文件中的表示形式混淆为JSON文本是很重要的。</p>
<p>JSON格式允许转义任何字符,而不仅仅是ASCII范围之外的Unicode字符:</p>
<pre><code>>>> import json
>>> json.loads(r'"\u0061"')
u'a'
>>> json.loads('"a"')
u'a'
</code></pre>
<p>不要把它与Python中的转义混淆,Python中的字符串文本用于源代码。<code>u"\u00f8"</code>是一个<em>单</em>Unicode字符,但是<code>"\u00f8"</code>在输出中是<em>八个</em>字符(在Python<em>源代码</em>中,可以将其改为<code>r'"\u00f8"' == '"\\u00f8"' == u'"\\u00f8"'</code>(反斜杠在Python<em>文本</em>和json文本中都是特殊的,可能会发生双转义</strong>)。JSON中也没有<code>\x</code>转义:</p>
<pre><code>>>> json.loads(r'"\x61"') # invalid JSON
Traceback (most recent call last):
...
ValueError: Invalid \escape: line 1 column 2 (char 1)
>>> r'"\x61"' # valid Python literal (6 characters)
'"\\x61"'
>>> '"\x61"' # valid Python literal with escape sequence (3 characters)
'"a"'
</code></pre>
<blockquote>
<p>The output of json.dumps() is a str, which is a byte string in Python 2. And thus shouldn't it escape characters as \xhh ?</p>
</blockquote>
<p><code>json.dumps(obj, ensure_ascii=True)</code>只生成可打印的ascii字符,因此<code>print repr(json.dumps(u"\xf8"))</code>将不包含用于表示</em>(<code>repr()</code>)不可打印字符(字节)的<code>\xhh</code>转义。</p>
<p><code>\u</code>即使对于仅限ascii的输入,也可能需要转义:</p>
<pre><code>#!/usr/bin/env python2
import json
print json.dumps(map(unichr, range(128)))
</code></pre>
<h3>输出</h3>
<pre><code>["\u0000", "\u0001", "\u0002", "\u0003", "\u0004", "\u0005", "\u0006", "\u0007",
"\b", "\t", "\n", "\u000b", "\f", "\r", "\u000e", "\u000f", "\u0010", "\u0011",
"\u0012", "\u0013", "\u0014", "\u0015", "\u0016", "\u0017", "\u0018", "\u0019",
"\u001a", "\u001b", "\u001c", "\u001d", "\u001e", "\u001f", " ", "!", "\"", "#",
"$", "%", "&", "'", "(", ")", "*", "+", ",", "-", ".", "/", "0", "1", "2", "3",
"4", "5", "6", "7", "8", "9", ":", ";", "<", "=", ">", "?", "@", "A", "B", "C",
"D", "E", "F", "G", "H", "I", "J", "K", "L", "M", "N", "O", "P", "Q", "R", "S",
"T", "U", "V", "W", "X", "Y", "Z", "[", "\\", "]", "^", "_", "`", "a", "b", "c",
"d", "e", "f", "g", "h", "i", "j", "k", "l", "m", "n", "o", "p", "q", "r", "s",
"t", "u", "v", "w", "x", "y", "z", "{", "|", "}", "~", "\u007f"]
</code></pre>
<blockquote>
<p>But isn't this quite confusing because \uxxxx is a unicode character and should be used inside a unicode string</p>
</blockquote>
<p><code>\uxxxx</code>是6个字符,在某些上下文(例如,在Python源代码中)中可以解释为单个字符<code>u"\uxxxx"</code>是一个Python文本,它在内存中创建一个带有单个Unicode字符的Unicode字符串。但是,如果在json文本中看到<code>\uxxxx</code>;如果加载单个Unicode字符(<code>json.loads()</code>),则<em>六个</em>字符可能表示该字符。</p>
<p>在这一点上,您应该理解为什么<code>len(json.loads('"\\\\"')) == 1</code>。</p>