Re: ICU4C API proposal - int64 support for NumberFormat

What about the C APIs in unum.h? Those need extension as well, along 
the lines of:

unum_formatInt64(...
unum_parseInt64(...

Deborah Goldsmith
Manager, Fonts / Unicode liaison
Apple Computer, Inc.
[email protected]

On Oct 31, 2003, at 3:13 PM, Doug Felt wrote:

> API change - add int64_t support to NumberFormat and associated 
> classes.
> See Jitterbug #813
> Please respond by Nov 7th.
>
> Note, there are two kinds of potential errors in user code that are 
> revealed by
> these changes (because they occurred in our tests)
>
> 1) Assuming that if a Formattable is a number and not a long, it is a 
> double.
> 2) Assuming that if a number is too large to fit in a long, then it 
> must be a
> double.
>
> The first occurred in a test that was converting all the formattables 
> to
> doubles.  It was testing the type of the formattable, and if it was 
> not kLong
> or kDouble, was returning an error.  Now there is a new type, kInt64, 
> which
> invalidated this old assumption.  It's a general problem when 
> extending a fixed
> range of types, people sometimes write code that doesn't rely on 
> positive
> identification of the type and instead relies on elimination, and this 
> fails.
>
> The second occurred in a test that was taking the maximum long value, 
> as a
> double and multiplying it by 1000.  This resulted in a double value 
> that
> represented an integer value.  The test tried to round trip this to 
> text and
> back, and just assumed (without checking) that the formattable must 
> contain a
> double, since it was 'too large' for the int representation.  This is 
> no longer
> true as an int64_t could hold the (integral) value of the original 
> double.
>
> Affected files:
>
> -tests-
> tsnmfmt.h
> numrgts.cpp
> numfmtst.cpp/h
>
> -i18n-
> fmtable.cpp/h
> numfmt.cpp/h
> digitlst.cpp/h (to be moved to i18n from common.  internal API)
> decimfmt.cpp/h
> choicfmt.cpp/h
>
> Affected APIs:
>
> -- Formattable--
> (Internally, stores int64_t instead of int32_t for its integral data.  
> All
> int32_t APIs remain and are augmented with similar int64_t APIs.)
>
>     /**
>      * Creates a Formattable object with an int64 number
>      * @param ll the int64 number.
>      * @draft ICU 2.8
>      */
>     Formattable(int64_t ll);
>
>     /**
>      * The list of possible data types of this Formattable object.
>      * @draft ICU 2.4
>      */
>     enum Type {
> 		/** @draft ICU 2.4 */
>         kDate,      // Date
> 		/** @draft ICU 2.4 */
>         kDouble,    // double
> 		/** @draft ICU 2.4 */
>         kLong,      // long
> 		/** @draft ICU 2.4 */
>         kString,    // UnicodeString
> 		/** @draft ICU 2.4 */
>         kArray,     // Formattable[]
> 		/** @draft ICU 2.8 */
>  		kInt64      // int64
>    };
>
>     /**
>      * Gets the int64 value of this object.
>      * @return    the int64 value of this object.
>      * @draft ICU 2.8
>      */
>     int64_t          getInt64(void) const;
>
>    /**
>      * Sets the int64 value of this object.
>      * @param ll    the new int64 value to be set.
>      * @draft ICU 2.8
>      */
>     void            setInt64(int64_t ll);
>
>
> -- NumberFormat --
> (APIs that took int32_t are overloaded to take int64_t.  Note, rather 
> than
> introducing
> a new pure virtual method overload, I gave the overload an 
> implementation so
> that
> existing user subclasses will still compile (we have such a class in 
> our
> tests).
>
>     /**
>      * Format an int64 number. These methods call the NumberFormat
>      * pure virtual format() methods with the default FieldPosition.
>      *
>      * @param number    The value to be formatted.
>      * @param appendTo  Output parameter to receive result.
>      *                  Result is appended to existing contents.
>      * @return          Reference to 'appendTo' parameter.
>      * @draft ICU 2.8
>      */
>     UnicodeString& format(  int64_t number,
>                             UnicodeString& appendTo) const;
>
>     /**
>      * Format an int64 number. (Not abstract to retain compatibility
>   * with earlier releases, however subclasses should override this
>   * method as it just delegates to format(int32_t number...);
>      *
>      * @param number    The value to be formatted.
>      * @param appendTo  Output parameter to receive result.
>      *                  Result is appended to existing contents.
>      * @param pos       On input: an alignment field, if desired.
>      *                  On output: the offsets of the alignment field.
>      * @return          Reference to 'appendTo' parameter.
>      * @draft ICU 2.8
>     */
>     virtual UnicodeString& format(int64_t number,
>                                   UnicodeString& appendTo,
>                                   FieldPosition& pos) const;
>
> -- DecimalFormat --
> (concrete implementation of NumberFormat)
>
>     /**
>      * Format an int64 number using base-10 representation.
>      *
>      * @param number    The value to be formatted.
>      * @param appendTo  Output parameter to receive result.
>      *                  Result is appended to existing contents.
>      * @param pos       On input: an alignment field, if desired.
>      *                  On output: the offsets of the alignment field.
>      * @return          Reference to 'appendTo' parameter.
>      * @draft ICU 2.8
>      */
>     virtual UnicodeString& format(int64_t number,
>                                   UnicodeString& appendTo,
>                                   FieldPosition& pos) const;
>
>     /**
>      * Redeclared NumberFormat method.
>      * Format an int64 number. These methods call the NumberFormat
>      * pure virtual format() methods with the default FieldPosition.
>      *
>      * @param number    The value to be formatted.
>      * @param appendTo  Output parameter to receive result.
>      *                  Result is appended to existing contents.
>      * @return          Reference to 'appendTo' parameter.
>      * @draft ICU 2.8
>      */
>     UnicodeString& format(int64_t number,
>                           UnicodeString& appendTo) const;
>
> -- ChoiceFormat --
> (This has to change since new overrides were added in NumberFormat, a 
> base
> class.
> Since ChoiceFormat's implementation is based on doubles, this has no
> interesting
> implementation.)
>
>     /**
>      * Format an int_64t number using this object's choices.
>      *
>      * @param number    The value to be formatted.
>      * @param appendTo  Output parameter to receive result.
>      *                  Result is appended to existing contents.
>      * @param pos       On input: an alignment field, if desired.
>      *                  On output: the offsets of the alignment field.
>      * @return          Reference to 'appendTo' parameter.
>      * @draft ICU 2.8
>      */
>     virtual UnicodeString& format(int64_t number,
>                                   UnicodeString& appendTo,
>                                   FieldPosition& pos) const;
>
>
> __________________________________
> Do you Yahoo!?
> Exclusive Video Premiere - Britney Spears
> http://launch.yahoo.com/promos/britneyspears/
> _______________________________________________
> icu mailing list
> [email protected]
> http://oss.software.ibm.com/developerworks/oss/mailman/listinfo/icu