improved documentation

Python/numpy style: Active description, start with capital, end with
full stop
This commit is contained in:
Martin Diehl 2020-09-08 06:55:04 +02:00
parent 43e9ea352a
commit 1f28592fcd
1 changed files with 11 additions and 11 deletions

View File

@ -101,7 +101,7 @@ end subroutine prec_init
!--------------------------------------------------------------------------------------------------
!> @brief equality comparison for float with double precision
!> @brief Test floating point numbers with double precision for equality.
! replaces "==" but for certain (relative) tolerance. Counterpart to dNeq
! https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/
! AlmostEqualRelative
@ -124,7 +124,7 @@ end function dEq
!--------------------------------------------------------------------------------------------------
!> @brief inequality comparison for float with double precision
!> @brief Test floating point numbers with double precision for inequality.
! replaces "!=" but for certain (relative) tolerance. Counterpart to dEq
! https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/
! AlmostEqualRelative NOT
@ -144,7 +144,7 @@ end function dNeq
!--------------------------------------------------------------------------------------------------
!> @brief equality to 0 comparison for float with double precision
!> @brief Test floating point number with double precision for equality to 0.
! replaces "==0" but everything not representable as a normal number is treated as 0. Counterpart to dNeq0
! https://de.mathworks.com/help/matlab/ref/realmin.html
! https://docs.oracle.com/cd/E19957-01/806-3568/ncg_math.html
@ -167,7 +167,7 @@ end function dEq0
!--------------------------------------------------------------------------------------------------
!> @brief inequality to 0 comparison for float with double precision
!> @brief Test floating point number with double precision for inequality to 0.
! replaces "!=0" but everything not representable as a normal number is treated as 0. Counterpart to dEq0
! https://de.mathworks.com/help/matlab/ref/realmin.html
! https://docs.oracle.com/cd/E19957-01/806-3568/ncg_math.html
@ -187,7 +187,7 @@ end function dNeq0
!--------------------------------------------------------------------------------------------------
!> @brief equality comparison for complex with double precision
!> @brief Test complex floating point numbers with double precision for equality.
! replaces "==" but for certain (relative) tolerance. Counterpart to cNeq
! https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/
! probably a component wise comparison would be more accurate than the comparsion of the absolute
@ -211,7 +211,7 @@ end function cEq
!--------------------------------------------------------------------------------------------------
!> @brief inequality comparison for complex with double precision
!> @brief Test complex floating point numbers with double precision for inequality.
! replaces "!=" but for certain (relative) tolerance. Counterpart to cEq
! https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/
! probably a component wise comparison would be more accurate than the comparsion of the absolute
@ -232,7 +232,7 @@ end function cNeq
!--------------------------------------------------------------------------------------------------
!> @brief interpret array of bytes (C_SIGNED_CHAR) as C_FLOAT (4 byte float)
!> @brief Decode byte array (C_SIGNED_CHAR) as C_FLOAT array (4 byte float).
!--------------------------------------------------------------------------------------------------
pure function bytes_to_C_FLOAT(bytes)
@ -246,7 +246,7 @@ end function bytes_to_C_FLOAT
!--------------------------------------------------------------------------------------------------
!> @brief interpret array of bytes (C_SIGNED_CHAR) as C_DOUBLE (8 byte float)
!> @brief Decode byte array (C_SIGNED_CHAR) as C_DOUBLE array (8 byte float).
!--------------------------------------------------------------------------------------------------
pure function bytes_to_C_DOUBLE(bytes)
@ -260,7 +260,7 @@ end function bytes_to_C_DOUBLE
!--------------------------------------------------------------------------------------------------
!> @brief interpret array of bytes (C_SIGNED_CHAR) as C_INT32_T (4 byte signed integer)
!> @brief Decode byte array (C_SIGNED_CHAR) as C_INT32_T array (4 byte signed integer).
!--------------------------------------------------------------------------------------------------
pure function bytes_to_C_INT32_T(bytes)
@ -274,7 +274,7 @@ end function bytes_to_C_INT32_T
!--------------------------------------------------------------------------------------------------
!> @brief interpret array of bytes (C_SIGNED_CHAR) as C_INT64_T (8 byte signed integer)
!> @brief Decode byte array (C_SIGNED_CHAR) as C_INT64_T array (8 byte signed integer).
!--------------------------------------------------------------------------------------------------
pure function bytes_to_C_INT64_T(bytes)
@ -288,7 +288,7 @@ end function bytes_to_C_INT64_T
!--------------------------------------------------------------------------------------------------
!> @brief check correctness of some prec functions
!> @brief Check correctness of some prec functions.
!--------------------------------------------------------------------------------------------------
subroutine selfTest