Difference between revisions of "EDT:EGL Language Primitive Type Conversions"

Revision as of 14:47, 19 July 2011

Please see the parent of this page, EDT:EGL Language.

Conversions Between Primitive Types (Table 4)

Conversion Matrix

To see if there's a conversion from one type to another, find the cell in "from" type's row and the "to" type's column.

This matrix might contain differences from RBD.

Key

blank cell = The types are incompatible.  Conversion is not supported.

a = Same types (not a conversion).

b = Booleans convert to strings as "true" or "false" (lowercase).

c = Strings convert to/from timestamps using a predefined format for timestamps (to be defined later).  The required fields come from the timestamp's pattern.   The conversions are not controlled by variables like RBD's defaultTimestampFormat.  Because of this, we can't convert from a string to a timestamp with no pattern.  We can convert from a timestamp with no pattern to a string.  Functions will be provided which convert to/from strings and timestamps with a format string supplied by the program.

d = Strings convert to/from numbers using our literal syntax for numbers.  The conversions are not controlled by variables like RBD's defaultNumericFormat.  Functions will be provided which convert to/from strings and numbers with a format string supplied by the program.

e = Strings convert to bytes using the bit pattern of the Unicode characters in the string.

f = All numeric types are compatible with each other, but some conversions can result in overflow or truncation.

g = Numeric types convert to bytes using the number's bit pattern. When the bytes type includes a length (i.e. "bytes(4)" not just "bytes") the conversion is legal only if the size of the number in bytes equals the length of the bytes type.

h = Conversion from bytes to numeric types is done by interpreting the bytes data as a value in the format of the numeric type.  If the bytes type includes a length (i.e. "bytes(4)" not just "bytes") the conversion is legal only if the size of the number in bytes equals the length of the bytes type.  If the bytes type doesn't include a length, the conversion will fail at run time if the size of the number in bytes is not equal to the length of the bytes data.

i = This is like h, with the additional rule that the decimal type must include a length.  (Without the decimal's length we don't know how to interpret the bytes data.)

j = Conversion is necessary if the "from" and "to" types both have parameters (i.e. "bytes(4)" not just "bytes"), and the type parameters aren't identical.

k = Strings convert to/from dates using a predefined format for dates (to be defined later). The conversions are not controlled by variables like RBD's defaultDateFormat. Functions will be provided which convert to/from strings and dates with a format string supplied by the program.

Notes on Table 4

1. We'll document error conditions of conversions, and their resulting exceptions. Whenever possible this documentation will be comments in the EGL source file for a type.
   
2. Our philosophy is that we can't make EDT be both efficient, and also behave the same in every environment when it comes to edge cases like numeric overflow. This means that in normal circumstances conversions to numeric types do not check for overflow: EGL does whatever the underlying language does in that situation. There will be special syntax for users to indicate that overflow checking is required. The syntax is To Be Determined. It may be an operation, something that looks like a function call where the source expression and target variable are passed in. What happens when an overflow is detected? We might throw an exception, but that's also To Be Determined.