Add a delimiters parameter to the csv.Sniffer constructor

Feature or enhancement

csv.Sniffer.sniff() accepts an optional delimiters argument which restricts the set of characters that can be guessed as a delimiter. has_header() has no such argument, so the restriction cannot be applied to it.

I propose to add an optional delimiters parameter to the constructor, so that it applies to the sniffer rather than to a single call:

sniffer = csv.Sniffer(',;\t')
if sniffer.has_header(sample):
    ...
dialect = sniffer.sniff(sample)

It is used by both sniff() and has_header(). The delimiters argument of sniff(), if given, still takes precedence.

The given delimiters are also preferred in that order if several of them split the sample equally well: csv.Sniffer(';,') guesses ';' where csv.Sniffer(',;') guesses ','.

Linked PRs

• gh-155060

先從 issue 中描述的 csv.Sniffer、sniff() 和 has_header() 進入點開始，然後檢查現有的 CSV 測試以了解其目前行為。完成的標準是：兩個方法都使用建構函式的 delimiters 限制，sniff() 可以覆寫該限制，而且相同的候選項遵循所提供的分隔符號順序。