`csv.writer` with `QUOTE_NONE` still requires non-emtpy `quotechar` and `escapechar`

[vadimkantorov]

Bug report

Bug description:

I'd like to output all quotes verbatim, without any escaping/quoting, but this fails:

import csv
w = csv.writer(open('test.tsv', 'w'), delimiter = '\t', quoting = csv.QUOTE_NONE)
w.writerow(('"hello"', '"world"'))
# _csv.Error: need to escape, but no escapechar set

When I set escapechar='' or quotechar='', it also fails:

import csv
w = csv.writer(open('test.tsv', 'w'), delimiter = '\t', quoting = csv.QUOTE_NONE, escapechar = '')
w.writerow(('"hello"', '"world"'))
# TypeError: "escapechar" must be a 1-character string

import csv
w = csv.writer(open('test.tsv', 'w'), delimiter = '\t', quoting = csv.QUOTE_NONE, quotechar = '')
w.writerow(('"hello"', '"world"'))
# TypeError: "quotechar" must be a 1-character string

I would suggest that under QUOTE_NONE, escaping of quotes should not be performed at all, or at least allow empty escapechar / quotechar

---

The workaround / hack I found:

import csv
w = csv.writer(open('test.tsv', 'w'), delimiter = '\t', quoting = csv.QUOTE_NONE, quotechar = '\t')
w.writerow(('"hello"', '"world"'))

CPython versions tested on:

3.12

Operating systems tested on:

Linux

Linked PRs

• gh-132813: Fix the csv documentation for quoting and escaping #133209
• gh-132813: Improve error messages for incorrect types and values of csv.Dialog attributes #133241
• [3.14] gh-132813: Improve error messages for incorrect types and values of csv.Dialog attributes (GH-133241) #135050

[picnixz]

The docs say:

Instructs writer objects to never quote fields. When the current delimiter occurs in output data it is preceded by the current escapechar character. If escapechar is not set, the writer will raise Error if any characters that require escaping are encountered.

So, it aligns with the expected behavior, namely:

If escapechar is not set, the writer will raise Error if any characters that require escaping are encountered.

I'm converting this into a feature request but since 3.14 is almost feature-frozen, it will need to wait 3.15, unless we consider this as a bug, which I don't as it's documented as such.

---

I don't think we need to change the meaning of QUOTE_NONE. However, we should allow a verbatim output, namely something that considers the content of a field as being whatever between two delimiters. If someone needs to use a delimiter inside a field, they must already have quoted it inside the .csv.

[vadimkantorov]

Yeah, I expected that QUOTE_NONE means "do nothing about quoting / attempt to escape quotes", but apparently it's not so...

[picnixz]

For that to happen, I think we need a new quoting strategy, something like QUOTE_IGNORE but this name is a bit confusing =/

[vadimkantorov]

new QUOTE_IGNORE might work! :)

[picnixz]

Yes, but the name is very ambiguous as it's too close to QUOTE_NONE. We need a name that convene the fact that we don't do anything but that is different from QUOTE_NONE. We need to think of a better name or think of an alternative (e.g., we could keep QUOTE_NONE and ignore quotes that required escapes with another option). I mainly used the csv module with the default settings, so I didn't encounter this and I don't know if it's not too niche in this case (and if it doesn't excess the scope of the module itself).

If we want to reduce the number of options, we could simply use flags (namely QUOTE_NONE | QUOTE_IGNORE_ERRORS) something like that?

[serhiy-storchaka]

Use quotechar=None.

The documentation is not accurate. It only mentions the current delimiter, but the quoting is also needed for quotechar, escapechar and the line separator.

[serhiy-storchaka]

On other hand, there may be a bug in the code too. In other places, quotechar is ignored when quoting is QUOTE_NONE, but not here.

I'm working on this.