@Schema maximum and minimum are Strings?

[EricWittmann]

For example, the @Schema annotation has this property:

https://github.com/eclipse/microprofile-open-api/blob/master/api/src/main/java/org/eclipse/microprofile/openapi/annotations/media/Schema.java#L113-L119

    /**
     * Sets the maximum numeric value for a property. 
     * Ignored if the value is an empty string.
     * 
     * @return the maximum value for this schema
     **/
    String maximum() default "";

The data model defines these properties as BigDecimal. Can I assume that this is done because java annotations don't support BigDecimal as a type for a property and that no other type would be sufficient?

[EricWittmann]

Note that the Schema model also has the property multipleOf which required a BigDecimal, but the corresponding property of the annotation is double.

[msavy]

Just a quick thought: could Hibernate's @Digits annotation serve as some inspiration? It has a variety of properties that allow you to constrain values in useful ways, such as @Digits(integer=10, fraction=2).

[EricWittmann]

Assigning to myself. :)

[EricWittmann]

OK so the reason that these properties are String instead of something more (seemingly) sensible (e.g. double) is that there is no default value that we could use to clearly indicate that the property was not included. In other words, if we made the property a double then there is no default value we could provide that is not also a valid value.

The way all these annotation properties work is that every one of them has a default value. When the annotation processor reads them, if the default value is found (for any given annotation property) then the processor is supposed to not include it in the final OAI document being generated. For String properties, the default value is typically empty string. So when the processor reads that property and sees empty string, it skips it.

For this particular property, switching to double would mean there is no sensible default that could indicate "skip it" to the processor.

We could do something like @Digits as @msavy suggests. I think that would allow us to use null as the "skip it" indicator. But it adds a fair bit of complexity for rather little value.

So for now I think we should update the javadoc to explain why the type is String and then close this issue.