Codec Base Classes¶

Thecodecs module defines a set of base classes which define theinterfaces for working with codec objects, and can also be used as the basisfor custom codec implementations.

Each codec has to define four interfaces to make it usable as codec in Python:stateless encoder, stateless decoder, stream reader and stream writer. Thestream reader and writers typically reuse the stateless encoder/decoder toimplement the file protocols. Codec authors also need to define how thecodec will handle encoding and decoding errors.

>>>'German ß, ♬'.encode(encoding='ascii',errors='backslashreplace')b'German \\xdf, \\u266c'>>>'German ß, ♬'.encode(encoding='ascii',errors='xmlcharrefreplace')b'German ß, ♬'

Encodings and Unicode¶

Strings are stored internally as sequences of code points inrangeU+0000–U+10FFFF. (SeePEP 393 formore details about the implementation.)Once a string object is used outside of CPU and memory, endiannessand how these arrays are stored as bytes become an issue. As with othercodecs, serialising a string into a sequence of bytes is known asencoding,and recreating the string from the sequence of bytes is known asdecoding.There are a variety of different text serialisation codecs, which arecollectivity referred to astext encodings.

The simplest text encoding (called'latin-1' or'iso-8859-1') mapsthe code points 0–255 to the bytes0x0–0xff, which means that a stringobject that contains code points aboveU+00FF can’t be encoded with thiscodec. Doing so will raise aUnicodeEncodeError that lookslike the following (although the details of the error message may differ):UnicodeEncodeError:'latin-1'codeccan'tencodecharacter'\u1234'inposition3:ordinalnotinrange(256).

There’s another group of encodings (the so called charmap encodings) that choosea different subset of all Unicode code points and how these code points aremapped to the bytes0x0–0xff. To see how this is done simply opene.g.encodings/cp1252.py (which is an encoding that is used primarily onWindows). There’s a string constant with 256 characters that shows you whichcharacter is mapped to which byte value.

All of these encodings can only encode 256 of the 1114112 code pointsdefined in Unicode. A simple and straightforward way that can store each Unicodecode point, is to store each code point as four consecutive bytes. There are twopossibilities: store the bytes in big endian or in little endian order. Thesetwo encodings are calledUTF-32-BE andUTF-32-LE respectively. Theirdisadvantage is that if e.g. you useUTF-32-BE on a little endian machine youwill always have to swap bytes on encoding and decoding.UTF-32 avoids thisproblem: bytes will always be in natural endianness. When these bytes are readby a CPU with a different endianness, then bytes have to be swapped though. Tobe able to detect the endianness of aUTF-16 orUTF-32 byte sequence,there’s the so called BOM (“Byte Order Mark”). This is the Unicode characterU+FEFF. This character can be prepended to everyUTF-16 orUTF-32byte sequence. The byte swapped version of this character (0xFFFE) is anillegal character that may not appear in a Unicode text. So when thefirst character in aUTF-16 orUTF-32 byte sequenceappears to be aU+FFFE the bytes have to be swapped on decoding.Unfortunately the characterU+FEFF had a second purpose asaZEROWIDTHNO-BREAKSPACE: a character that has no width and doesn’t allowa word to be split. It can e.g. be used to give hints to a ligature algorithm.With Unicode 4.0 usingU+FEFF as aZEROWIDTHNO-BREAKSPACE has beendeprecated (withU+2060 (WORDJOINER) assuming this role). NeverthelessUnicode software still must be able to handleU+FEFF in both roles: as a BOMit’s a device to determine the storage layout of the encoded bytes, and vanishesonce the byte sequence has been decoded into a string; as aZEROWIDTHNO-BREAKSPACE it’s a normal character that will be decoded like any other.

There’s another encoding that is able to encode the full range of Unicodecharacters: UTF-8. UTF-8 is an 8-bit encoding, which means there are no issueswith byte order in UTF-8. Each byte in a UTF-8 byte sequence consists of twoparts: marker bits (the most significant bits) and payload bits. The marker bitsare a sequence of zero to four1 bits followed by a0 bit. Unicode characters areencoded like this (with x being payload bits, which when concatenated give theUnicode character):

The least significant bit of the Unicode character is the rightmost x bit.

As UTF-8 is an 8-bit encoding no BOM is required and anyU+FEFF character inthe decoded string (even if it’s the first character) is treated as aZEROWIDTHNO-BREAKSPACE.

Without external information it’s impossible to reliably determine whichencoding was used for encoding a string. Each charmap encoding candecode any random byte sequence. However that’s not possible with UTF-8, asUTF-8 byte sequences have a structure that doesn’t allow arbitrary bytesequences. To increase the reliability with which a UTF-8 encoding can bedetected, Microsoft invented a variant of UTF-8 (that Python calls"utf-8-sig") for its Notepad program: Before any of the Unicode charactersis written to the file, a UTF-8 encoded BOM (which looks like this as a bytesequence:0xef,0xbb,0xbf) is written. As it’s rather improbablethat any charmap encoded file starts with these byte values (which would e.g.map to

LATIN SMALL LETTER I WITH DIAERESIS

RIGHT-POINTING DOUBLE ANGLE QUOTATION MARK

INVERTED QUESTION MARK

in iso-8859-1), this increases the probability that autf-8-sig encoding can becorrectly guessed from the byte sequence. So here the BOM is not used to be ableto determine the byte order used for generating the byte sequence, but as asignature that helps in guessing the encoding. On encoding the utf-8-sig codecwill write0xef,0xbb,0xbf as the first three bytes to the file. Ondecodingutf-8-sig will skip those three bytes if they appear as the firstthree bytes in the file. In UTF-8, the use of the BOM is discouraged andshould generally be avoided.

Standard Encodings¶

Python comes with a number of codecs built-in, either implemented as C functionsor with dictionaries as mapping tables. The following table lists the codecs byname, together with a few common aliases, and the languages for which theencoding is likely used. Neither the list of aliases nor the list of languagesis meant to be exhaustive. Notice that spelling alternatives that only differ incase or use a hyphen instead of an underscore are also valid aliases; therefore,e.g.'utf-8' is a valid alias for the'utf_8' codec.

Many of the character sets support the same languages. They vary in individualcharacters (e.g. whether the EURO SIGN is supported or not), and in theassignment of characters to code positions. For the European languages inparticular, the following variants typically exist:

• an ISO 8859 codeset

• a Microsoft Windows code page, which is typically derived from an 8859 codeset,but replaces control characters with additional graphic characters

• an IBM EBCDIC code page

• an IBM PC code page, which is ASCII compatible

Python Specific Encodings¶

A number of predefined codecs are specific to Python, so their codec names haveno meaning outside Python. These are listed in the tables below based on theexpected input and output types (note that while text encodings are the mostcommon use case for codecs, the underlying codec infrastructure supportsarbitrary data transforms rather than just text encodings). For asymmetriccodecs, the stated meaning describes the encoding direction.

encodings.idna — Internationalized Domain Names in Applications¶

This module implementsRFC 3490 (Internationalized Domain Names inApplications) andRFC 3492 (Nameprep: A Stringprep Profile forInternationalized Domain Names (IDN)). It builds upon thepunycode encodingandstringprep.

If you need the IDNA 2008 standard fromRFC 5891 andRFC 5895, use thethird-partyidna module.

These RFCs together define a protocol to support non-ASCII characters in domainnames. A domain name containing non-ASCII characters (such aswww.Alliancefrançaise.nu) is converted into an ASCII-compatible encoding(ACE, such aswww.xn--alliancefranaise-npb.nu). The ACE form of the domainname is then used in all places where arbitrary characters are not allowed bythe protocol, such as DNS queries, HTTPHost fields, and soon. This conversion is carried out in the application; if possible invisible tothe user: The application should transparently convert Unicode domain labels toIDNA on the wire, and convert back ACE labels to Unicode before presenting themto the user.

Python supports this conversion in several ways: theidna codec performsconversion between Unicode and ACE, separating an input string into labelsbased on the separator characters defined insection 3.1 of RFC 3490and converting each label to ACE as required, and conversely separating an inputbyte string into labels based on the. separator and converting any ACElabels found into unicode. Furthermore, thesocket moduletransparently converts Unicode host names to ACE, so that applications need notbe concerned about converting host names themselves when they pass them to thesocket module. On top of that, modules that have host names as functionparameters, such ashttp.client andftplib, accept Unicode hostnames (http.client then also transparently sends an IDNA hostname in theHost field if it sends that field at all).

When receiving host names from the wire (such as in reverse name lookup), noautomatic conversion to Unicode is performed: applications wishing to presentsuch host names to the user should decode them to Unicode.

The moduleencodings.idna also implements the nameprep procedure, whichperforms certain normalizations on host names, to achieve case-insensitivity ofinternational domain names, and to unify similar characters. The nameprepfunctions can be used directly if desired.

encodings.idna.nameprep(label)¶

Return the nameprepped version oflabel. The implementation currently assumesquery strings, soAllowUnassigned is true.

encodings.idna.ToASCII(label)¶

Convert a label to ASCII, as specified inRFC 3490.UseSTD3ASCIIRules isassumed to be false.

encodings.idna.ToUnicode(label)¶

Convert a label to Unicode, as specified inRFC 3490.

encodings.mbcs — Windows ANSI codepage¶

This module implements the ANSI codepage (CP_ACP).

encodings.utf_8_sig — UTF-8 codec with BOM signature¶

This module implements a variant of the UTF-8 codec. On encoding, a UTF-8 encodedBOM will be prepended to the UTF-8 encoded bytes. For the stateful encoder thisis only done once (on the first write to the byte stream). On decoding, anoptional UTF-8 encoded BOM at the start of the data will be skipped.