[8.5.2] Error handling padCharacter in EwNumericFormat>>#formatNumber:

In most places, EwNumericFormat handles padCharacter correctly, using #nextPut: to add it to the stream.
However, I just noticed one place where it incorrectly uses #nextPutAll:.

        formatChar == $S ifTrue: [
            self isSignFixed ifTrue: [outputStream nextPutAll: self signString].
            self isSignFloating ifTrue: [outputStream nextPutAll: self padCharacter]].

The effect is that the padding does not get emitted. Characters answer 0 for their size.

On Wednesday, October 1, 2014 1:09:22 PM UTC-7, Seth Berman wrote:

I created case 56368 for this item.

At first glance, looks like previousFormatChar in the 'formatChar == $T' will always be nil due to its scope so I will just make note to review the whole method.

Thanks. I think the latter is good. The loop schematic is:

    self formatString doWithIndex: [:formatChar :i |

        previousFormatChar := formatChar].

I read the condition as "if already writing digits or the padded leading zeros, emit a thousands separator otherwise a padding character."

On Wednesday, October 1, 2014 2:16:24 PM UTC-7, Seth Berman wrote:

In your version is previousFormatChar declared within the inner loop?

I see something like.

self formatString doWithIndex: [:formatChar :i |
        | previousFormatChar |

        "test the value of previousFormatChar (which I think is always nil)"

        previousFormatChar := formatChar].

Do you see something different?

Oops. That is what I see, alright. That ain't gonna work.