Type of strings and bytes inconsistent iterating manually or in .tolist()

[lgray]

check = ak.Array([{"x": "b"}, {"x": "b"}, {"x": "a"}, {"x": "b"}])
print(check.layout)

print(ak.tolist(check))
print(ak.tolist(check.layout))

results in:

<RecordArray>
    <type>{"x": string}</type>
    <field index="0" key="x">
        <ListOffsetArray64>
            <type>string</type>
            <offsets><Index64 i="[0 1 2 3 4]" offset="0" at="0x7fb342145600"/></offsets>
            <content><NumpyArray format="B" shape="4" data="0x 62626162" at="0x7fb342034800">
                <type>utf8</type>
            </NumpyArray></content>
        </ListOffsetArray64>
    </field>
</RecordArray>
[{'x': [98]}, {'x': [98]}, {'x': [97]}, {'x': [98]}]
[{'x': [98]}, {'x': [98]}, {'x': [97]}, {'x': [98]}]

check = ak.Array([{"x": b"b"}, {"x": b"b"}, {"x": b"a"}, {"x": b"b"}])
print(check.layout)

print(ak.tolist(check))
print(ak.tolist(check.layout))

results in:

<RecordArray>
    <type>{"x": bytes}</type>
    <field index="0" key="x">
        <ListOffsetArray64>
            <type>bytes</type>
            <offsets><Index64 i="[0 1 2 3 4]" offset="0" at="0x7fb3431d6e00"/></offsets>
            <content><NumpyArray format="B" shape="4" data="0x 62626162" at="0x7fb34314e200">
                <type>byte</type>
            </NumpyArray></content>
        </ListOffsetArray64>
    </field>
</RecordArray>
[{'x': [98]}, {'x': [98]}, {'x': [97]}, {'x': [98]}]
[{'x': [98]}, {'x': [98]}, {'x': [97]}, {'x': [98]}]

However:

for out in check:
    print(out["x"])

results in (similar for strings):

b'b'
b'b'
b'a'
b'b'

but:

for out in check.layout:
    print(out["x"])

results in

<NumpyArray format="B" shape="1" data="0x 62" at="0x7fb3439c8200">
    <type>byte</type>
</NumpyArray>
<NumpyArray format="B" shape="1" data="0x 62" at="0x7fb3439c8200">
    <type>byte</type>
</NumpyArray>
<NumpyArray format="B" shape="1" data="0x 61" at="0x7fb3439c8200">
    <type>byte</type>
</NumpyArray>
<NumpyArray format="B" shape="1" data="0x 62" at="0x7fb3439c8200">
    <type>byte</type>
</NumpyArray>

It looks like the underlying behavior is correct but the handling is inconsistent?

[lgray]

The shortest reproducer is:

check = ak.Array([{"x": b"b"}, {"x": b"b"}, {"x": b"a"}, {"x": b"b"}])
print(check)
print(ak.tolist(check))
print(ak.tolist(ak.fromiter(ak.tolist(check))))

which gives:

[{x: b'b'}, {x: b'b'}, {x: b'a'}, {x: b'b'}]
[{'x': [98]}, {'x': [98]}, {'x': [97]}, {'x': [98]}]
[{'x': [98]}, {'x': [98]}, {'x': [97]}, {'x': [98]}]