Overall it looks great but please check travis has an error:
/tarantool/src/lib/core/decimal.c:319:8: error: unused variable ‘tmp’ [-
Werror=unused-variable]
char *tmp = (char *)decPackedFromNumber((uint8_t *)data, len, &scale, dec);

Perhaps, it was caused by assert elimination because of the Release-type 
build.

On Tuesday, June 4, 2019 7:04:17 PM MSK Serge Petrenko wrote:
> Add fixed-point decimal type to tarantool core.
> Adapt decNumber floating-point decimal library for the purpose, write a
> small wrapper and add unit tests.
> 
> A new decimal type is an alias for decNumber numbers from the decNumber
> library.
> Arithmetic operations (+, -, *, /) and some mathematic functions
> (ln, log10, exp, pow, sqrt) are available.
> 
> We introduce a single context for all the arithmetic operations
> on decimals, which enforces both number precision and scale to be
> in range [0, 38].
> 
> Every mathematic function has precision as one of its parameters, and
> the results are rounded to fit the desired precision.
> 
> Part of #692
> ---
>  CMakeLists.txt              |   7 +
>  cmake/BuildDecNumber.cmake  |  13 ++
>  src/CMakeLists.txt          |   1 +
>  src/lib/core/CMakeLists.txt |   3 +-
>  src/lib/core/decimal.c      | 293 ++++++++++++++++++++++++++++++++++++
>  src/lib/core/decimal.h      | 159 +++++++++++++++++++
>  test/unit/CMakeLists.txt    |   2 +
>  test/unit/decimal.c         | 168 +++++++++++++++++++++
>  test/unit/decimal.result    |  51 +++++++
>  9 files changed, 696 insertions(+), 1 deletion(-)
>  create mode 100644 cmake/BuildDecNumber.cmake
>  create mode 100644 src/lib/core/decimal.c
>  create mode 100644 src/lib/core/decimal.h
>  create mode 100644 test/unit/decimal.c
>  create mode 100644 test/unit/decimal.result
> 
> diff --git a/CMakeLists.txt b/CMakeLists.txt
> index 7658fc6c9..bfb15effb 100644
> --- a/CMakeLists.txt
> +++ b/CMakeLists.txt
> @@ -431,6 +431,13 @@ else()
>      find_package(MsgPuck)
>  endif()
> 
> +#
> +# decNumber
> +#
> +
> +include(BuildDecNumber)
> +decnumber_build()
> +
>  #
>  # LibYAML
>  #
> diff --git a/cmake/BuildDecNumber.cmake b/cmake/BuildDecNumber.cmake
> new file mode 100644
> index 000000000..80942fe05
> --- /dev/null
> +++ b/cmake/BuildDecNumber.cmake
> @@ -0,0 +1,13 @@
> +#
> +# A macro to build the bundled decNumber lisbrary.
> +macro(decnumber_build)
> +    set(decnumber_src
> +	${PROJECT_SOURCE_DIR}/third_party/decNumber/decNumber.c
> +	${PROJECT_SOURCE_DIR}/third_party/decNumber/decContext.c
> +    )
> +
> +    add_library(decNumber STATIC ${decnumber_src})
> +
> +    set(DECNUMBER_INCLUDE_DIR ${PROJECT_BINARY_DIR}/third_party/decNumber)
> +    unset(decnumber_src)
> +endmacro(decnumber_build)
> diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
> index 183cc04ef..9d196755f 100644
> --- a/src/CMakeLists.txt
> +++ b/src/CMakeLists.txt
> @@ -13,6 +13,7 @@ include_directories(${MSGPUCK_INCLUDE_DIRS})
>  include_directories(${CURL_INCLUDE_DIRS})
>  include_directories(${ICU_INCLUDE_DIRS})
>  include_directories(${ICONV_INCLUDE_DIRS})
> +include_directories(${DECNUMBER_INCLUDE_DIR})
> 
>  set(LIBUTIL_FREEBSD_SRC ${CMAKE_SOURCE_DIR}/third_party/libutil_freebsd)
>  include_directories(${LIBUTIL_FREEBSD_SRC})
> diff --git a/src/lib/core/CMakeLists.txt b/src/lib/core/CMakeLists.txt
> index eb10b11c3..66e430a25 100644
> --- a/src/lib/core/CMakeLists.txt
> +++ b/src/lib/core/CMakeLists.txt
> @@ -26,6 +26,7 @@ set(core_sources
>      trigger.cc
>      mpstream.c
>      port.c
> +    decimal.c
>  )
> 
>  if (TARGET_OS_NETBSD)
> @@ -37,7 +38,7 @@ endif()
> 
>  add_library(core STATIC ${core_sources})
> 
> -target_link_libraries(core salad small uri ${LIBEV_LIBRARIES}
> +target_link_libraries(core salad small uri decNumber ${LIBEV_LIBRARIES}
>                        ${LIBEIO_LIBRARIES} ${LIBCORO_LIBRARIES}
>                        ${MSGPUCK_LIBRARIES})
> 
> diff --git a/src/lib/core/decimal.c b/src/lib/core/decimal.c
> new file mode 100644
> index 000000000..12250da95
> --- /dev/null
> +++ b/src/lib/core/decimal.c
> @@ -0,0 +1,293 @@
> +/*
> + * Copyright 2019, Tarantool AUTHORS, please see AUTHORS file.
> + *
> + * Redistribution and use in source and binary forms, with or
> + * without modification, are permitted provided that the following
> + * conditions are met:
> + *
> + * 1. Redistributions of source code must retain the above
> + *    copyright notice, this list of conditions and the
> + *    following disclaimer.
> + *
> + * 2. Redistributions in binary form must reproduce the above
> + *    copyright notice, this list of conditions and the following
> + *    disclaimer in the documentation and/or other materials
> + *    provided with the distribution.
> + *
> + * THIS SOFTWARE IS PROVIDED BY <COPYRIGHT HOLDER> ``AS IS'' AND
> + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
> + * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
> + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL
> + * <COPYRIGHT HOLDER> OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
> + * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
> + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
> + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
> + * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
> + * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
> + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF
> + * THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
> + * SUCH DAMAGE.
> + */
> +
> +#include "decimal.h"
> +#include "third_party/decNumber/decContext.h"
> +#include <stdlib.h>
> +#include <assert.h>
> +
> +#define MAX(a, b) ((a) > (b) ? (a) : (b))
> +#define MIN(a, b) ((a) > (b) ? (b) : (a))
> +
> +/** A single context for all the arithmetic operations. */
> +static decContext decimal_context = {
> +	/* Maximum precision during operations. */
> +	TARANTOOL_MAX_DECIMAL_DIGITS,
> +	/*
> +	 * Maximum decimal lagarithm of the number.
> +	 * Allows for precision = TARANTOOL_MAX_DECIMAL_DIGITS
> +	 */
> +	TARANTOOL_MAX_DECIMAL_DIGITS - 1,
> +	/*
> +	 * Minimal adjusted exponent. The smallest absolute value will be
> +	 * exp((1 - TARANTOOL_MAX_DECIMAL_DIGITS) - 1) =
> +	 * exp(-TARANTOOL_MAX_DECIMAL_DIGITS) allowing for scale =
> +	 * TARANTOOL_MAX_DECIMAL_DIGITS
> +	 */
> +	-1,
> +	/* Rounding mode: .5 rounds away from 0. */
> +	DEC_ROUND_HALF_UP,
> +	/* Turn off signalling for failed operations. */
> +	0,
> +	/* Status holding occured events. Initially empty. */
> +	0,
> +	/* Turn off exponent clamping. */
> +	0
> +};
> +
> +/**
> + * Return operation status and clear it for future checks.
> + *
> + * @return 0 if ok, bitwise or of decNumber errors, if any.
> + */
> +static inline uint32_t
> +decimal_get_op_status(decContext *context)
> +{
> +	uint32_t status = decContextGetStatus(context);
> +	decContextZeroStatus(&decimal_context);
> +	/*
> +	 * Clear warnings. Every value less than 0.1 is
> +	 * subnormal, DEC_Inexact and DEC_Rounded result
> +	 * from rounding. DEC_Inexact with DEC_Subnormal
> +	 * together result in DEC_Underflow. DEC_Clamped
> +	 * happens after underflow if rounding to zero.
> +	 */
> +	return status & ~(uint32_t)(DEC_Inexact | DEC_Rounded | 
DEC_Underflow |
> +				    DEC_Subnormal | DEC_Clamped);
> +}
> +
> +/**
> + * A finalizer for all the operations.
> + * Check the operation context status and empty it.
> + *
> + * @return NULL if finalization failed.
> + *         result pointer otherwise.
> + */
> +static inline decimal *
> +decimal_finalize(decimal *res, decContext *context)
> +{
> +	uint32_t status = decimal_get_op_status(context);
> +	if (status || ! decNumberIsFinite(res)) {
> +		return NULL;
> +	}
> +	return res;
> +}
> +
> +uint8_t decimal_precision(const decimal *dec) {
> +	return dec->exponent <= 0 ? MAX(dec->digits, -dec->exponent) :
> +				    dec->digits + dec->exponent;
> +}
> +
> +uint8_t decimal_scale(const decimal *dec) {
> +	return dec->exponent < 0 ? -dec->exponent : 0;
> +}
> +
> +decimal *
> +decimal_zero(decimal *dec)
> +{
> +	decNumberZero(dec);
> +	return dec;
> +}
> +
> +decimal *
> +decimal_from_string(decimal *dec, const char *str)
> +{
> +	decNumberFromString(dec, str, &decimal_context);
> +	return decimal_finalize(dec, &decimal_context);
> +}
> +
> +decimal *
> +decimal_from_int(decimal *dec, int32_t num)
> +{
> +	decNumberFromInt32(dec, num);
> +	return decimal_finalize(dec, &decimal_context);
> +}
> +
> +decimal *
> +decimal_from_uint(decimal *dec, uint32_t num)
> +{
> +	decNumberFromUInt32(dec, num);
> +	return decimal_finalize(dec, &decimal_context);
> +}
> +
> +char *
> +decimal_to_string(const decimal *dec, char *buf)
> +{
> +	return decNumberToString(dec, buf);
> +}
> +
> +int32_t
> +decimal_to_int(const decimal *dec)
> +{
> +	decNumber res;
> +	decNumberToIntegralValue(&res, dec, &decimal_context);
> +	return decNumberToInt32(&res, &decimal_context);
> +}
> +
> +uint32_t
> +decimal_to_uint(const decimal *dec)
> +{
> +	decNumber res;
> +	decNumberToIntegralValue(&res, dec, &decimal_context);
> +	return decNumberToUInt32(&res, &decimal_context);
> +}
> +
> +int
> +decimal_compare(const decimal *lhs, const decimal *rhs)
> +{
> +	decNumber res;
> +	decNumberCompare(&res, lhs, rhs, &decimal_context);
> +	return decNumberToInt32(&res, &decimal_context);
> +}
> +
> +decimal *
> +decimal_abs(decimal *res, const decimal *dec)
> +{
> +	decNumberAbs(res, dec, &decimal_context);
> +	return res;
> +}
> +
> +decimal *
> +decimal_add(decimal *res, const decimal *lhs, const decimal *rhs)
> +{
> +	decNumberAdd(res, lhs, rhs, &decimal_context);
> +
> +	return decimal_finalize(res, &decimal_context);
> +}
> +
> +decimal *
> +decimal_sub(decimal *res, const decimal *lhs, const decimal *rhs)
> +{
> +	decNumberSubtract(res, lhs, rhs, &decimal_context);
> +
> +	return decimal_finalize(res, &decimal_context);
> +}
> +
> +decimal *
> +decimal_mul(decimal *res, const decimal *lhs, const decimal *rhs)
> +{
> +	decNumberMultiply(res, lhs, rhs, &decimal_context);
> +
> +	return decimal_finalize(res, &decimal_context);
> +}
> +
> +decimal *
> +decimal_div(decimal *res, const decimal *lhs, const decimal *rhs)
> +{
> +	decNumberDivide(res, lhs, rhs, &decimal_context);
> +
> +	return decimal_finalize(res, &decimal_context);
> +}
> +
> +decimal *
> +decimal_log10(decimal *res, const decimal *lhs, uint32_t precision)
> +{
> +	if (precision >= TARANTOOL_MAX_DECIMAL_DIGITS)
> +		return NULL;
> +
> +	/* A separate context to enforce desired precision. */
> +	static decContext op_context;
> +	op_context.digits = precision;
> +	op_context.emax = op_context.digits - 1;
> +	op_context.emin = -1;
> +
> +	decNumberLog10(res, lhs, &op_context);
> +
> +	return decimal_finalize(res, &op_context);
> +}
> +
> +decimal *
> +decimal_ln(decimal *res, const decimal *lhs, uint32_t precision)
> +{
> +	if (precision >= TARANTOOL_MAX_DECIMAL_DIGITS)
> +		return NULL;
> +
> +	/* A separate context to enforce desired precision. */
> +	static decContext op_context;
> +	op_context.digits = precision;
> +	op_context.emax = op_context.digits - 1;
> +	op_context.emin = -1;
> +
> +	decNumberLn(res, lhs, &op_context);
> +
> +	return decimal_finalize(res, &op_context);
> +}
> +
> +decimal *
> +decimal_pow(decimal *res, const decimal *lhs, const decimal *rhs, uint32_t
> precision) +{
> +	if (precision >= TARANTOOL_MAX_DECIMAL_DIGITS)
> +		return NULL;
> +
> +	/* A separate context to enforce desired precision. */
> +	static decContext op_context;
> +	op_context.digits = precision;
> +	op_context.emax = op_context.digits - 1;
> +	op_context.emin = -1;
> +
> +	decNumberPower(res, lhs, rhs, &op_context);
> +
> +	return decimal_finalize(res, &op_context);
> +}
> +
> +decimal *
> +decimal_exp(decimal *res, const decimal *lhs, uint32_t precision)
> +{
> +	if (precision >= TARANTOOL_MAX_DECIMAL_DIGITS)
> +		return NULL;
> +
> +	/* A separate context to enforce desired precision. */
> +	static decContext op_context;
> +	op_context.digits = precision;
> +	op_context.emax = op_context.digits - 1;
> +	op_context.emin = -1;
> +
> +	decNumberExp(res, lhs, &op_context);
> +
> +	return decimal_finalize(res, &op_context);
> +}
> +
> +decimal *
> +decimal_sqrt(decimal *res, const decimal *lhs, uint32_t precision)
> +{
> +	if (precision >= TARANTOOL_MAX_DECIMAL_DIGITS)
> +		return NULL;
> +
> +	/* A separate context to enforce desired precision. */
> +	static decContext op_context;
> +	op_context.digits = precision;
> +	op_context.emax = op_context.digits - 1;
> +	op_context.emin = -1;
> +
> +	decNumberSquareRoot(res, lhs, &op_context);
> +
> +	return decimal_finalize(res, &op_context);
> +}
> diff --git a/src/lib/core/decimal.h b/src/lib/core/decimal.h
> new file mode 100644
> index 000000000..f2812f500
> --- /dev/null
> +++ b/src/lib/core/decimal.h
> @@ -0,0 +1,159 @@
> +#ifndef TARANTOOL_LIB_CORE_DECIMAL_H_INCLUDED
> +#define TARANTOOL_LIB_CORE_DECIMAL_H_INCLUDED
> +/*
> + * Copyright 2019, Tarantool AUTHORS, please see AUTHORS file.
> + *
> + * Redistribution and use in source and binary forms, with or
> + * without modification, are permitted provided that the following
> + * conditions are met:
> + *
> + * 1. Redistributions of source code must retain the above
> + *    copyright notice, this list of conditions and the
> + *    following disclaimer.
> + *
> + * 2. Redistributions in binary form must reproduce the above
> + *    copyright notice, this list of conditions and the following
> + *    disclaimer in the documentation and/or other materials
> + *    provided with the distribution.
> + *
> + * THIS SOFTWARE IS PROVIDED BY <COPYRIGHT HOLDER> ``AS IS'' AND
> + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
> + * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
> + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL
> + * <COPYRIGHT HOLDER> OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
> + * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
> + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
> + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
> + * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
> + * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
> + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF
> + * THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
> + * SUCH DAMAGE.
> + */
> +#define TARANTOOL_MAX_DECIMAL_DIGITS 38
> +#define DECNUMDIGITS TARANTOOL_MAX_DECIMAL_DIGITS
> +#define DECSTRING_NO_EXPONENT
> +#include "third_party/decNumber/decNumber.h"
> +
> +typedef decNumber decimal;
> +
> +/**
> + * @return decimal precision,
> + * i.e. the amount of decimal digits in
> + * its representation.
> + */
> +uint8_t
> +decimal_precision(const decimal *dec);
> +
> +/**
> + * @return decimal scale,
> + * i.e. the number of decimal digits after
> + * the decimal separator.
> + */
> +uint8_t
> +decimal_scale(const decimal *dec);
> +
> +/**
> + * Initialize a zero decimal number.
> + */
> +decimal *
> +decimal_zero(decimal *dec);
> +
> +/**
> + * Initialize a decimal with a value from the string.
> + *
> + * If the number is less, than 10^TARANTOOL_MAX_DECIMAL_DIGITS,
> + * but has excess digits in fractional part, it will be rounded.
> + *
> + * @return NULL if string is invalid or
> + * the number is too big (>= 10^TARANTOOL_MAX_DECIMAL_DIGITS)
> + */
> +decimal *
> +decimal_from_string(decimal *dec, const char *str);
> +
> +/**
> + * Initialize a decimal with an integer value.
> + *
> + * @return NULL if precicion is insufficient to hold
> + * the value or precision/scale are out of bounds.
> +*/
> +decimal *
> +decimal_from_int(decimal *dec, int32_t num);
> +
> +/** @copydoc decimal_from_int */
> +decimal *
> +decimal_from_uint(decimal *dec, uint32_t num);
> +
> +/**
> + * Write the decimal to a string.
> + * A string has to be at least dec->digits + 3 bytes in size.
> + */
> +char *
> +decimal_to_string(const decimal *dec, char *buf);
> +
> +/**
> + * Cast decimal to an integer value. The number will be rounded
> + * if it has a fractional part.
> + */
> +int32_t
> +decimal_to_int(const decimal *dec);
> +
> +/** @copydoc decimal_to_int */
> +uint32_t
> +decimal_to_uint(const decimal *dec);
> +
> +/**
> + * Compare 2 decimal values.
> + * @return -1, lhs < rhs,
> + *	    0, lhs = rhs,
> + *	    1, lhs > rhs
> + */
> +int
> +decimal_compare(const decimal *lhs, const decimal *rhs);
> +
> +/**
> + * res is set to the absolute value of dec
> + * decimal_abs(&a, &a) is allowed.
> + */
> +decimal *
> +decimal_abs(decimal *res, const decimal *dec);
> +
> +/*
> + * Arithmetic ops: add, subtract, multiply and divide.
> + * Return result pointer on success, NULL on an error (overflow).
> + */
> +
> +decimal *
> +decimal_add(decimal *res, const decimal *lhs, const decimal *rhs);
> +
> +decimal *
> +decimal_sub(decimal *res, const decimal *lhs, const decimal *rhs);
> +
> +decimal *
> +decimal_mul(decimal *res, const decimal *lhs, const decimal *rhs);
> +
> +decimal *
> +decimal_div(decimal *res, const decimal *lhs, const decimal *rhs);
> +
> +/*
> + * log10, ln, pow, exp, sqrt.
> + * Calculate the appropriate function and then round the result
> + * to the desired precision.
> + * Return result pointer on success, NULL on an error (overflow).
> + */
> +decimal *
> +decimal_log10(decimal *res, const decimal *lhs, uint32_t precision);
> +
> +decimal *
> +decimal_ln(decimal *res, const decimal *lhs, uint32_t precision);
> +
> +decimal *
> +decimal_pow(decimal *res, const decimal *lhs, const decimal *rhs, uint32_t
> precision); +
> +decimal *
> +decimal_exp(decimal *res, const decimal *lhs, uint32_t precision);
> +
> +decimal *
> +decimal_sqrt(decimal *res, const decimal *lhs, uint32_t precision);
> +
> +#endif /* TARANTOOL_LIB_CORE_DECIMAL_H_INCLUDED */
> diff --git a/test/unit/CMakeLists.txt b/test/unit/CMakeLists.txt
> index 70be6366c..e3de34b16 100644
> --- a/test/unit/CMakeLists.txt
> +++ b/test/unit/CMakeLists.txt
> @@ -68,6 +68,8 @@ add_executable(vclock.test vclock.cc)
>  target_link_libraries(vclock.test vclock unit)
>  add_executable(xrow.test xrow.cc)
>  target_link_libraries(xrow.test xrow unit)
> +add_executable(decimal.test decimal.c)
> +target_link_libraries(decimal.test core unit)
> 
>  add_executable(fiber.test fiber.cc)
>  set_source_files_properties(fiber.cc PROPERTIES COMPILE_FLAGS -O0)
> diff --git a/test/unit/decimal.c b/test/unit/decimal.c
> new file mode 100644
> index 000000000..adfbead91
> --- /dev/null
> +++ b/test/unit/decimal.c
> @@ -0,0 +1,168 @@
> +#include "unit.h"
> +#include "decimal.h"
> +#include <limits.h>
> +#include <string.h>
> +
> +int
> +main(void)
> +{
> +	plan(50);
> +
> +	char buf[TARANTOOL_MAX_DECIMAL_DIGITS + 3];
> +	char buf2[TARANTOOL_MAX_DECIMAL_DIGITS + 3];
> +	char *b = "2.718281828";
> +
> +	decimal s;
> +	decimal *ret;
> +
> +	ret = decimal_from_string(&s, b);
> +	isnt(ret, NULL, "Basic construction from string.");
> +	decimal_to_string(&s, buf);
> +	is(strcmp(buf, b), 0, "Correct construction and to_string 
conversion.");
> +
> +	ret = decimal_from_int(&s, INT_MAX);
> +	decimal_to_string(&s, buf);
> +	sprintf(buf2, "%d", INT_MAX);
> +	is(strcmp(buf, buf2), 0, "Correct construction from INT_MAX.");
> +	is(decimal_to_int(&s), INT_MAX, "Simple conversion back to int 
INT_MAX");
> +
> +	ret = decimal_from_int(&s, INT_MIN);
> +	decimal_to_string(&s, buf);
> +	sprintf(buf2, "%d", INT_MIN);
> +	is(strcmp(buf, buf2), 0, "Correct construction from INT_MIN.");
> +	is(decimal_to_int(&s), INT_MIN, "Simple conversion bact to int 
INT_MIN");
> +
> +	char *up1 = "2.5";
> +	char *down1 = "2.49";
> +	decimal_from_string(&s, up1);
> +	is(decimal_to_int(&s), 3, ".5 Rounds up");
> +	decimal_from_string(&s, down1);
> +	is(decimal_to_int(&s), 2, ".49 Rounds down");
> +
> +	char *l = "1.23456789123456789123456789123456789123";
> +	ret = decimal_from_string(&s, l);
> +	isnt(ret, NULL, "Precision too high. Rounding happens.");
> +	ok(decimal_precision(&s) == 38 && decimal_scale(&s) == 37 , 
"Construction
> is correct."); +	char *ll = 
"123456789123456789123456789123456789123";
> +	ret = decimal_from_string(&s, ll);
> +	is(ret, NULL, "Precision too high and scale = 0. Cannot round");
> +
> +	/* 38 digits. */
> +	char *long_str = "0.0000000000000000000000000000000000001";
> +	ret = decimal_from_string(&s, long_str);
> +	isnt(ret, NULL, "Construncting the smallest possible number from 
string");
> +	decimal_to_string(&s, buf);
> +	is(strcmp(buf, long_str), 0, "Correct representation of smallest 
possible
> number"); +
> +	/* Comparsions. */
> +	char *max_str = "3.11", *min_str = "3.0999";
> +	decimal max, min;
> +	decimal_from_string(&max, max_str);
> +	decimal_from_string(&min, min_str);
> +	is(decimal_compare(&max, &min), 1, "max > min");
> +	is(decimal_compare(&min, &max), -1, "min < max");
> +	is(decimal_compare(&max, &max), 0, "max == max");
> +
> +	ret = decimal_from_string(&s, "-3.456");
> +	isnt(ret, NULL, "Construction from negative numbers");
> +	decimal_to_string(&s, buf);
> +	is(strcmp(buf, "-3.456"), 0, "Correct construction for negatives");
> +	ret = decimal_abs(&s, &s);
> +	isnt(ret, NULL, "Abs");
> +	decimal_to_string(&s, buf);
> +	is(strcmp(buf, "3.456"), 0, "Correct abs");
> +
> +	/* Arithmetic ops. */
> +	decimal d, check;
> +	ret = decimal_from_string(&s, b);
> +	ret = decimal_from_string(&d, "1.25");
> +	sprintf(buf2, "%.9f", 1.25 + 2.718281828);
> +	ret = decimal_add(&d, &d, &s);
> +	isnt(ret, NULL, "Simple addition");
> +	decimal_to_string(&d, buf);
> +	is(strcmp(buf, buf2), 0, "Simple addition is correct");
> +
> +	ret = decimal_sub(&d, &d, &s);
> +	isnt(ret, NULL, "Simple subtraction");
> +	decimal_from_string(&check, "1.25");
> +	is(decimal_compare(&d, &check), 0, "Simple subtraction is 
correct");
> +
> +	decimal_from_int(&s, 4);
> +	ret = decimal_mul(&s, &s, &d);
> +	isnt(ret, NULL, "Simple multiplication");
> +	decimal_from_string(&check, "5.0");
> +	is(decimal_compare(&s, &check), 0 , "Simple multiplication is 
correct");
> +
> +	ret = decimal_div(&s, &s, &d);
> +	isnt(ret, NULL, "Simple division");
> +	decimal_from_string(&check, "4.0");
> +	is(decimal_compare(&s, &check), 0, "Simple division is correct");
> +
> +	/* Math. */
> +	ret = decimal_from_string(&s, "40.96");
> +	ret = decimal_from_string(&check, "6.4");
> +	ret = decimal_sqrt(&s, &s, 2);
> +	isnt(ret, NULL, "sqrt");
> +	is(decimal_compare(&s, &check), 0, "sqrt is correct");
> +
> +	ret = decimal_from_string(&s, "40.96");
> +	ret = decimal_from_string(&d, "0.5");
> +	ret = decimal_pow(&s, &s, &d, 2);
> +	isnt(ret, NULL, "pow");
> +	is(decimal_compare(&s, &check), 0, "pow is correct");
> +
> +	ret = decimal_from_string(&s, "3.000");
> +	ret = decimal_exp(&d, &s, 4);
> +	isnt(ret, NULL, "exp");
> +
> +	ret = decimal_from_string(&check, "20.09");
> +	is(decimal_compare(&d, &check), 0, "exp is correct")
> +	ret = decimal_ln(&d, &d, 4);
> +	isnt(ret, NULL, "ln");
> +	/*
> +	 * ln is wrong by 1 unit in last position.
> +	 * (3.001 instead of 3.000, 3.0001 instead of 3.0000 and so on)
> +	 */
> +	decimal_from_string(&s, "3.001");
> +	is(decimal_compare(&d, &s), 0, "ln is correct");
> +
> +				      /* 10^3.5 */
> +	ret = decimal_from_string(&s, "3162.27766");
> +	ret = decimal_log10(&d, &s, 2);
> +	isnt(ret, NULL, "log10");
> +	ret = decimal_from_string(&check, "3.5");
> +	is(decimal_compare(&d, &check), 0, "log10 is correct");
> +
> +	/* Advanced test. */
> +	/* 38 digits. */
> +	char *bignum = "33.333333333333333333333333333333333333";
> +	char *test =   "133.33333333333333333333333333333333333";
> +	ret = decimal_from_string(&s, bignum);
> +	ret = decimal_from_int(&d, 4);
> +	ret = decimal_mul(&s, &s, &d);
> +	isnt(ret, NULL, "Rounding when more than 
TARANTOOL_MAX_DECIMAL_DIGITS
> digits"); +	ret = decimal_from_string(&check, test);
> +	is(decimal_compare(&s, &check), 0, "Rounding is correct");
> +	is(decimal_precision(&s), 38, "Correct precision");
> +	is(decimal_scale(&s), 35, "Correct scale");
> +
> +	char *small = "0.000000000000000000000000001";
> +	ret = decimal_from_string(&s, small);
> +	ret = decimal_mul(&s, &s, &s);
> +	isnt(ret, NULL, "Rounding too small number to zero");
> +	ret = decimal_from_int(&check, 0);
> +	is(decimal_compare(&s, &check), 0, "Rounding is correct");
> +	is(decimal_precision(&s), 38, "Correct precision");
> +	is(decimal_scale(&s), 38, "Correct scale");
> +
> +	decimal_from_string(&s, small);
> +	decimal_from_string(&d, "10000000000000000000");
> +	ret = decimal_div(&s, &s, &d);
> +	isnt(ret, NULL, "Rounding too small number to zero");
> +	is(decimal_compare(&s, &check), 0, "Rounding is correct");
> +	decimal_to_string(&s, buf);
> +	is(decimal_precision(&s), 38, "Correct precision");
> +	is(decimal_scale(&s), 38, "Correct scale");
> +
> +	check_plan();
> +}
> diff --git a/test/unit/decimal.result b/test/unit/decimal.result
> new file mode 100644
> index 000000000..02bee118a
> --- /dev/null
> +++ b/test/unit/decimal.result
> @@ -0,0 +1,51 @@
> +1..50
> +ok 1 - Basic construction from string.
> +ok 2 - Correct construction and to_string conversion.
> +ok 3 - Correct construction from INT_MAX.
> +ok 4 - Simple conversion back to int INT_MAX
> +ok 5 - Correct construction from INT_MIN.
> +ok 6 - Simple conversion bact to int INT_MIN
> +ok 7 - .5 Rounds up
> +ok 8 - .49 Rounds down
> +ok 9 - Precision too high. Rounding happens.
> +ok 10 - Construction is correct.
> +ok 11 - Precision too high and scale = 0. Cannot round
> +ok 12 - Construncting the smallest possible number from string
> +ok 13 - Correct representation of smallest possible number
> +ok 14 - max > min
> +ok 15 - min < max
> +ok 16 - max == max
> +ok 17 - Construction from negative numbers
> +ok 18 - Correct construction for negatives
> +ok 19 - Abs
> +ok 20 - Correct abs
> +ok 21 - Simple addition
> +ok 22 - Simple addition is correct
> +ok 23 - Simple subtraction
> +ok 24 - Simple subtraction is correct
> +ok 25 - Simple multiplication
> +ok 26 - Simple multiplication is correct
> +ok 27 - Simple division
> +ok 28 - Simple division is correct
> +ok 29 - sqrt
> +ok 30 - sqrt is correct
> +ok 31 - pow
> +ok 32 - pow is correct
> +ok 33 - exp
> +ok 34 - exp is correct
> +ok 35 - ln
> +ok 36 - ln is correct
> +ok 37 - log10
> +ok 38 - log10 is correct
> +ok 39 - Rounding when more than TARANTOOL_MAX_DECIMAL_DIGITS digits
> +ok 40 - Rounding is correct
> +ok 41 - Correct precision
> +ok 42 - Correct scale
> +ok 43 - Rounding too small number to zero
> +ok 44 - Rounding is correct
> +ok 45 - Correct precision
> +ok 46 - Correct scale
> +ok 47 - Rounding too small number to zero
> +ok 48 - Rounding is correct
> +ok 49 - Correct precision
> +ok 50 - Correct scale