The stdlib_linalg_state
module provides a derived type holding information on the state of linear algebra operations, and procedures for expert control of linear algebra workflows.
All linear algebra procedures are engineered to support returning an optional linalg_state_type
variable to hold such information, as a form of expert API. If the user does not require state information but fatal errors are encountered during the execution of linear algebra routines, the program will undergo a hard stop.
Instead, if the state argument is present, the program will never stop but will return detailed error information into the state handler.
linalg_state_type
derived typeThe linalg_state_type
is an extension of the state_type
derived type, containing an integer error flag and fixed-size character strings to store an error message and the location of the error state change.
Fixed-size string storage was chosen to facilitate the compiler's memory allocation and ultimately ensure maximum computational performance.
A similarly named generic interface, linalg_state_type
, is provided to allow the developer to create diagnostic messages and raise error flags easily. The call starts with an error flag or the location of the event and is followed by an arbitrary list of integer
, real
, complex
, or character
variables. Numeric variables may be provided as either scalars or rank-1 (array) inputs.
The following convenience type-bound procedures are inherited from state_type
and available:
- print()
returns an allocatable character string containing state location, message, and error flag;
- print_message()
returns an allocatable character string containing the state message;
- ok()
returns a logical
flag that is .true.
in case of successful state (flag==LINALG_SUCCESS
);
- error()
returns a logical
flag that is .true.
in case of an error state (flag/=LINALG_SUCCESS
).
Experimental
program example_state1
use stdlib_linalg_state, only: linalg_state_type, LINALG_SUCCESS, LINALG_VALUE_ERROR, &
operator(/=)
implicit none
type(linalg_state_type) :: err
! To create a state variable, we enter its integer state flag, followed by a list of variables
! that will be automatically assembled into a formatted error message. No need to provide string formats
err = linalg_state_type(LINALG_VALUE_ERROR,'just an example with scalar ',&
'integer=',1,'real=',2.0,'complex=',(3.0,1.0),'and array ',[1,2,3],'inputs')
! Print flag
print *, err%print()
! Check success
print *, 'Check error: ',err%error()
print *, 'Check flag : ',err /= LINALG_SUCCESS
end program example_state1
The module provides the following state flags, mapped to the general state_type
error flags:
- LINALG_SUCCESS
: Successful execution (equivalent to STDLIB_SUCCESS
)
- LINALG_VALUE_ERROR
: Numerical errors (such as infinity, not-a-number, range bounds) are encountered (equivalent to STDLIB_VALUE_ERROR
).
- LINALG_ERROR
: Linear Algebra errors are encountered, such as non-converging iterations, and impossible operations (equivalent to STDLIB_LINALG_ERROR
).
- LINALG_INTERNAL_ERROR
: Provided as a developer safeguard for internal errors that should never occur (equivalent to STDLIB_INTERNAL_ERROR
).
The module provides overloaded comparison operators for all comparisons of a linalg_state_type
variable with an integer error flag: <
, <=
, ==
, >=
, >
, /=
.