Thread: Same double precision operations, different results
Please can someone tell me why this behavior? Thx!!
I know it's calculating over coordinates on same points, but the operations of values (all double precision) differs even forcing with cast
(*) Here some double precision calculations and obtaining values
# Select
# cos(radians(39.9826557)) as f1,
# cos(radians(39.9826557)) as f2,
# cos(radians(-0.04773120000004383) - radians(-0.04773120000004383)) as f3,
# sin(radians(39.9826557)) as f4,
# sin(radians(39.9826557)) as f5;
f1 | f2 | f3 | f4 | f5
-------------------+-------------------+----+-------------------+-------------------
0.766238989559398 | 0.766238989559398 | 1 | 0.642555686986733 | 0.642555686986733
(1 row)
________________________________________
(*) Here above same obtaining values involved in simple operations
# select (0.766238989559398 * 0.766238989559398 * 1 + 0.642555686986733 * 0.642555686986733) calc;
calc
----------------------------------
0.999999999999999633651488135693
(1 row)
________________________________________
(*) Here above queries playing together ---> Result = 1 ????? NOOOO! It should be 0.999999999999999633651488135693
# select (v.f1 * v.f2 * v.f3 + v.f4 * v.f5) calc from ( select
cos(radians(39.9826557)) as f1,
cos(radians(39.9826557)) as f2,
cos(radians(-0.04773120000004383) - radians(-0.04773120000004383)) as f3,
sin(radians(39.9826557)) as f4,
sin(radians(39.9826557)) as f5 ) as v;
calc
------
1
(1 row)
________________________________________
(*) Yep, this produces problems with acos. Here acos working directly with the above values. OK this is like acos(0.999999999999999633651488135693)
# select acos(0.766238989559398 * 0.766238989559398 * 1 + 0.642555686986733 * 0.642555686986733);
acos
----------------------
2.58095682795179e-08
(1 row)
________________________________________
(*) Yep, you can see same result
# select acos(0.999999999999999633651488135693);
acos
----------------------
2.58095682795179e-08
(1 r
________________________________________
(*) ok acos values from -1 to 1 you can see acos(1) here
# select acos(1);
acos
------
0
(1 row)
________________________________________
(*) Why this calculation produces 1 and not 0.999999999999999633651488135693 ?????????
# select (
cos(radians(39.9826557))
* cos(radians(39.9826557))
* cos(radians(-0.04773120000004383) - radians(-0.04773120000004383))
+ sin(radians(39.9826557))
* sin(radians(39.9826557))
) calc;
calc
------
1
(1 row)
________________________________________
(*) any way, the result is 1, then if acos(1) get results, this must get same result .... NOPS!! "input is out of range"
# select acos(
# cos(radians(39.9826557))
# * cos(radians(39.9826557))
# * cos(radians(-0.04773120000004383) - radians(-0.04773120000004383))
# + sin(radians(39.9826557))
# * sin(radians(39.9826557))
# );
ERROR: input is out of range
________________________________________
(*) You are thinking: it's some cast problem ... NOPS!! "input is out of range"
# select acos( cast (
cos(radians(39.9826557))
* cos(radians(39.9826557))
* cos(radians(-0.04773120000004383) - radians(-0.04773120000004383))
+ sin(radians(39.9826557))
* sin(radians(39.9826557)) as double precision )
);
ERROR: input is out of range
Really, I can get the reason!!! Some idea? Workaround? Thx
Daniel Vázquez <daniel2d2art@gmail.com> writes:
> Please can someone tell me why this behavior? Thx!!
You're confusing numeric and double precision; in particular this
calculation is *not* being done in float8, but numeric:
> # select (0.766238989559398 * 0.766238989559398 * 1 + 0.642555686986733 *
> 0.642555686986733) calc;
> (*) Why this calculation produces 1 and not 0.999999999999999633651488135693
Actually, it's not producing 1, but a smidgen more:
regression=# set extra_float_digits TO 3;
SET
regression=# select ( cast ( cos(radians(39.9826557)) * cos(radians(39.9826557)) * cos(radians(-0.04773120000004383)
-radians(-0.04773120000004383)) + sin(radians(39.9826557)) * sin(radians(39.9826557)) as double precision ));
float8
---------------------1.00000000000000022
(1 row)
You've got roundoff error either way, but this way happens to be in the
direction that makes acos() complain.
regards, tom lane
<div dir="ltr">Thx Tom!!<div style="style"><br /></div><div style="style">Yep, I focused on a<span style="color:rgb(0,0,0);font-family:verdana,sans-serif;font-size:12px;line-height:18.234375px">lltrigonometric functionstake arguments and return values of type </span><tt class="" style="font-size:12px;color:rgb(0,0,0);line-height:18.234375px">doubleprecision</tt><span style="color:rgb(0,0,0);font-family:verdana,sans-serif;font-size:12px;line-height:18.234375px">.Looking for the error I lostfocus on numeric values directly on the select and diff inner calculations. </span></div><div style="style"><span style="color:rgb(0,0,0);font-family:verdana,sans-serif;font-size:12px;line-height:18.234375px"><br/></span></div><div style="style"><spanstyle="color:rgb(0,0,0);font-family:verdana,sans-serif;font-size:12px;line-height:18.234375px">I thinkbest approach will be maintain double precision on trigonometric calculations for faster and cast to numeric beforeacos operation.</span></div><div style="style"><span style="color:rgb(0,0,0);font-family:verdana,sans-serif;font-size:12px;line-height:18.234375px"><br/></span></div><div style="style"><spanstyle="color:rgb(0,0,0);font-family:verdana,sans-serif;font-size:12px;line-height:18.234375px">Thank youman!</span></div><div style="style"><span style="color:rgb(0,0,0);font-family:verdana,sans-serif;font-size:12px;line-height:18.234375px"><br/></span></div><div style="style"><spanstyle="color:rgb(0,0,0);font-family:verdana,sans-serif;font-size:12px;line-height:18.234375px"><br /></span></div></div><divclass="gmail_extra"><br /><br /><div class="gmail_quote">2014-02-13 18:26 GMT+01:00 Tom Lane <spandir="ltr"><<a href="mailto:tgl@sss.pgh.pa.us" target="_blank">tgl@sss.pgh.pa.us</a>></span>:<br /><blockquoteclass="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="">DanielVázquez <<a href="mailto:daniel2d2art@gmail.com">daniel2d2art@gmail.com</a>> writes:<br /> > Pleasecan someone tell me why this behavior? Thx!!<br /><br /></div>You're confusing numeric and double precision; in particularthis<br /> calculation is *not* being done in float8, but numeric:<br /><div class=""><br /> > # select (0.766238989559398* 0.766238989559398 * 1 + 0.642555686986733 *<br /> > 0.642555686986733) calc;<br /><br /></div><divclass="">> (*) Why this calculation produces 1 and not 0.999999999999999633651488135693<br /><br /></div>Actually,it's not producing 1, but a smidgen more:<br /><br /> regression=# set extra_float_digits TO 3;<br /> SET<br/> regression=# select ( cast (<br /><div class=""> cos(radians(39.9826557))<br /> * cos(radians(39.9826557))<br/> * cos(radians(-0.04773120000004383) - radians(-0.04773120000004383))<br /> + sin(radians(39.9826557))<br/> * sin(radians(39.9826557)) as double precision )<br /> );<br /></div> float8<br />---------------------<br /> 1.00000000000000022<br /> (1 row)<br /><br /> You've got roundoff error either way, but thisway happens to be in the<br /> direction that makes acos() complain.<br /><br /> regards, tomlane<br /></blockquote></div><br /></div>