The previous section described Data sources and Queries which provide raw data. Raw data can be used as is, or can be processed further for the purposes of the report. This is where expressions come in.

There are three main data types in OpenCReports: numeric, string, and datetime. OpenCReports automatically detects the type of raw data supplied by Queries for SQL and some file based Data sources, and the application can supply extra data to indicate column data types.

Expressions can use and result in any of these types.

A numeric expression's result is a number. It uses high numeric precision. Most functions and operators deal with numbers.

For historic record, RLIB was designed for the US and for slower computers. It used a fixed point numeric representation. For the US, with its strong currency and prices expressed in low numbers, this was an acceptable design decision at the time. But for countries, where currencies are a few orders of magnitude weaker and conversely, the prices are similarly higher, the fixed point numeric value range was easily overflown (especially in report variables that added values), leading to wrong data in the report output.

Another potential problem with fixed point numeric representation is that converting numbers from the input data to this internal representation always rounds down. The numeric error (i.e. the difference between floating point values and fixed point values) can be demonstrated even with small data sets that add up percentages.

OpenCReports uses high precision floating point values. Technically, it's 256-bit precision GNU MPFR numerics by default, and the precision can be modified by the application if needed. This allows handling very large and very small numbers and directly consuming the SQL numeric and decimal types or arbitrary precision, or using bcmath numerics in PHP. This also allows scientific computation or monetary calculations even with late stage hyperinflation prices.

A string expression's result is arbitrary text. Strings can be concatenated or otherwise processed by string functions.

A datetime expression may store a date, a time (with or without timezone) or both. Also, it may store a time interval, e.g. 2 months that can be added to or subtracted from another datetime value.

There is a further data type: error. Errors usually occur if there is an error in processing, e.g. when a function argument does not match its expected parameter data type. The error type is a special case of strings: it stores a string literal, the error message. As opposed to the string data type, an error cannot be processed further by passing them as function arguments or operator operands. Instead, the first error is propagated up from subexpressions to the final result of the expression.