NeoCSV nil-Handling

I'd like a hook to specify default values more dynamically. The use case is
that I created a glue layer between Magritte and NeoCSV so one can
meta-describe how to extract CSV fields into objects.

For example:
```
descriptionDate
        <magritteDescription>
       
        ^ MADateDescription new
                        accessor: #date;
                        priority: 400;
                        propertyAt: #csvFieldName put: 'Date';
                        propertyAt: #csvReader put: [ :s | (s, '/2017') asDate ];
                        default: Date today;
                        yourself
```

The problem is that there doesn't seem to be a way to plug #default: above
into NeoCSV because other hooks like #addField:converter: seem to be skipped
for nil values, shortcutting to a statically described per-csv-file
emptyFieldValue.



-----
Cheers,
Sean
--
Sent from: http://forum.world.st/Pharo-Smalltalk-Developers-f1294837.html

Sean,

The initial idea of the empty value (defaulting to nil) and the short circuiting of further conversion, is precisely to help in the writing of the converter blocks, so that

 [ :s | s asDate ]

will never receive nil or an empty string and consequently does not have to test on it.

I suppose I could imagine an option like #passEmptyFieldsToConverters, which would allow (and force you) to write converter blocks like

 [ :s | s isEmptyOrNil ifTrue: [ Date doomsDay ] ifFalse: [ s asDate ] ]

so that you can specify your own custom empty value per field.
What do you think ?

Sven
 

Sven Van Caekenberghe-2 wrote
> What do you think ?

Could #emptyFieldValue: accept a block instead of the actual value? Or be
per-field? Or both? Then the default could still be set via the current
mechanism, keeping the usual API simple, but the user would have a bit more
control.



-----
Cheers,
Sean
--
Sent from: http://forum.world.st/Pharo-Smalltalk-Developers-f1294837.html

Sean,

> On 19 Feb 2018, at 02:33, Sean P. DeNigris <[hidden email]> wrote:
>
> Sven Van Caekenberghe-2 wrote
>> What do you think ?
>
> Could #emptyFieldValue: accept a block instead of the actual value? Or be
> per-field? Or both? Then the default could still be set via the current
> mechanism, keeping the usual API simple, but the user would have a bit more
> control.

Have a look at the following:

===
Name: Neo-CSV-Core-SvenVanCaekenberghe.27
Author: SvenVanCaekenberghe
Time: 19 February 2018, 10:28:24.688684 pm
UUID: a08d6c1c-f921-0d00-aaa4-29b9044088b7
Ancestors: Neo-CSV-Core-SvenVanCaekenberghe.26

Add special option #passNil for NeoCSVReader>>#emptyFieldValue: so that converter blocks do get nil passed in for empty/missing fields and can act accordingly

Add NeoCSVReaderTests>>#testEmptyFieldValuePassNil
===
Name: Neo-CSV-Tests-SvenVanCaekenberghe.24
Author: SvenVanCaekenberghe
Time: 19 February 2018, 10:28:03.52381 pm
UUID: 1598291b-f921-0d00-aaa3-c7c0044088b7
Ancestors: Neo-CSV-Tests-SvenVanCaekenberghe.23

Add special option #passNil for NeoCSVReader>>#emptyFieldValue: so that converter blocks do get nil passed in for empty/missing fields and can act accordingly

Add NeoCSVReaderTests>>#testEmptyFieldValuePassNil
===

Now you can do as follows:

| date1 date2 |
date1 := '1900-01-01' asDate.
date2 := '2000-12-31' asDate.
self
  assert: ((NeoCSVReader on: 'date1,date2\2018-01-01,2018-02-01\2018-01-01,\,2018-02-01\\' withCRs readStream)
    emptyFieldValue: #passNil;
    addFieldConverter: [ :input | input ifNil: [ date1 ] ifNotNil: [ input asDate ] ];
    addFieldConverter: [ :input | input ifNil: [ date2 ] ifNotNil: [ input asDate ] ];
    skipHeader;
    upToEnd)
equals: (Array
    with: (Array with: '2018-01-01' asDate with: '2018-02-01' asDate)
    with: (Array with: '2018-01-01' asDate with: date2)
    with: (Array with: date1 with: '2018-02-01' asDate)
    with: (Array with: date1 with: date2)).

Note how the first field's default empty value is now date1 and second field's is date2.

Sven