Request: Despite `dump`'s `width` parameter, don't create inline lists/dicts when they are flat with a single item

[AndydeCleyre]

Hi, and thanks as always!

While the resulting line is shorter than the width limit, it's without the tradeoff of fewer lines.

I'm requesting that either this behavior change, or that a new parameter be offered, to forgo inlining when the structure has a length of 1 and there is no nesting (either a list with one item which is a string, or a dict with one entry whose value is a string).

Another way of putting it: don't use the inline form when the alternative is already a single line.

Current behavior:

>>> from nestedtext import dumps
>>> print(dumps({'mode': 'plain'}, width=20))
{mode: plain}

>>> print(dumps({'navigation': {'mode': 'plain'}}, width=20))
navigation:
    {mode: plain}

>>> print(dumps({'mode': [1]}, width=20))
{mode: [1]}

>>> print(dumps([1], width=20))
[1]

>>> print(dumps([1, 2], width=20))
[1, 2]

>>> print(dumps([{'mode': 'one'}], width=20))
[{mode: one}]

Requested behavior:

>>> from nestedtext import dumps
>>> print(dumps({'mode': 'plain'}, width=20))
mode: plain

>>> print(dumps({'navigation': {'mode': 'plain'}}, width=20))
navigation:
    mode: plain

>>> print(dumps({'mode': [1]}, width=20))
{mode: [1]}

>>> print(dumps([1], width=20))
- 1

>>> print(dumps([1, 2], width=20))
[1, 2]

>>> print(dumps([{'mode': 'one'}], width=20))
[{mode: one}]

Diff of current and requested behavior from above:

--- current.py	2026-01-27 13:49:06.581031210 -0500
+++ requested.py	2026-01-27 13:48:56.801759721 -0500
@@ -1,16 +1,16 @@
 >>> from nestedtext import dumps
 >>> print(dumps({'mode': 'plain'}, width=20))
-{mode: plain}
+mode: plain
 
 >>> print(dumps({'navigation': {'mode': 'plain'}}, width=20))
 navigation:
-    {mode: plain}
+    mode: plain
 
 >>> print(dumps({'mode': [1]}, width=20))
 {mode: [1]}
 
 >>> print(dumps([1], width=20))
-[1]
+- 1
 
 >>> print(dumps([1, 2], width=20))
 [1, 2]

Side by side view of the same diff:

[KenKundert]

I kind of like the idea, but I am not sold on it. It makes the output simpler in some cases, but it can also make it inconsistent and somewhat unpredictable. For example:

>>> import nestedtext as nt

>>> data = [
...     {},
...     {0:0},
...     {0:0, 1:1},
...     {0:0, 1:1, 2:2},
... ]
>>> print(nt.dumps(data, width=90, inline_level=1))
-
    {}
-
    0: 0
-
    {0: 0, 1: 1}
-
    {0: 0, 1: 1, 2: 2}

or

>>> data = [
...     [],
...     [0],
...     [0, 1],
...     [0, 1, 2],
... ]
>>> print(nt.dumps(data, width=90, inline_level=1))
-
    []
-
    - 0
-
    [0, 1]
-
    [0, 1, 2]

The logic is somewhat complex, which can make the result unpredictable from the user's perspective.

[AndydeCleyre]

I hear you. What do you think about an additional parameter specifying the minimum line length for a structure to be eligible for inlining? In my case I'd set that value to 2. The default would be 1, not changing the current behavior.

[KenKundert]

What do you think about an additional parameter specifying the minimum line length for a structure to be eligible for inlining?

Do you mean width in characters or depth?

[AndydeCleyre]

Oh sorry, I meant line count.

[KenKundert]

I am going to put this off for a while. I am thinking of including it in a larger enhancement, but given my current schedule, it will take some time to get to.