Sphere output precision limited to 15 significant digits #118
Comments
Oh, that's a good point. |
|
Dear All, I have tried to create a binary backup using the pg_dumpbinary software, hoping that it would keep the original data with the needed precision, but the following error appears: ERROR: no binary output function available for type public.spoly Thus, going back to the quick solution for the text output, I guess that in fact the method set_sphere_output_precision(n) (which is limited to n<=DBL_DIG, i.e. 15) tries to imitate the PostgreSQL behavior of "SET extra_float_digits". Regards. |
|
The value changed from 15 to 17 in PG12. |
|
Dear All, Obviously, there should be a way to set 17 digits for output. But I'm not sure about DBL_DECIMAL_DIG. It seems to appear in C11. May be to utilize extra_float_digits? Let me, please, dig into this topic. I will come with some proposal later. |
|
Some other datatypes (like my Perhaps it is just a matter of using |
|
I agree, we should unify this behaviour with postgresql. What I'm concerned is the backward compatibility. I propose:
If user decides to use postgresql extra_float_digits, set_sphere_output_precision shouldn't be called or called with -1, for example. It shouldn't affect old scripts with set_sphere_output_precision. What do you think? P.S. set_sphere_output_precision seems to be absent in the doc. Some clarification should be written in the doc. |
|
That all makes sense to me, @vitcpp. Thanks! |
pgSphere used its own setting (set_sphere_output_precision function) to limit the output precision of double values, that could not be greater than 15 significant digits (DBL_DIG). It introduced some problems with dump/restore. PostgreSQL uses its own precision setting: extra_float_digits. The PostgreSQL setting allows to use more significant digits. This patch changes the default pgSphere output behaviour to utilize PostgreSQL extra_float_digits. Now, extra_float_digits is used by default, until set_sphere_output_precision is called. The old behaviour is kept for compatibility purposes. Once, set_sphere_output_precision is called, pgSphere starts to use the old behaviour (read, please, issue postgrespro#118 discussion). The patch introduces a new function - reset_sphere_output_precision. It is used to reset to the PostgreSQL behaviour after using set_sphere_output_precision.
pgSphere used its own setting (set_sphere_output_precision function) to limit the output precision of double values, that could not be greater than 15 significant digits (DBL_DIG). It introduced some problems with dump/restore. PostgreSQL uses its own precision setting: extra_float_digits. The PostgreSQL setting allows to use more significant digits. This patch changes the default pgSphere output behaviour to utilize PostgreSQL extra_float_digits. Now, extra_float_digits is used by default, until set_sphere_output_precision is called. The old behaviour is kept for compatibility purposes. Once, set_sphere_output_precision is called, pgSphere starts to use the old behaviour (read, please, issue postgrespro#118 discussion). The patch introduces a new function - reset_sphere_output_precision. It is used to reset to the PostgreSQL behaviour after using set_sphere_output_precision.
Fix output precision limit for double values (issue #118) pgSphere used its own setting (set_sphere_output_precision function) to limit the output precision of double values, that could not be greater than 15 significant digits (DBL_DIG). It introduced some problems with dump/restore. PostgreSQL uses its own precision setting: extra_float_digits. The PostgreSQL setting allows to use more significant digits. This patch changes the default pgSphere output behaviour to utilize PostgreSQL extra_float_digits. Now, extra_float_digits is used by default, until set_sphere_output_precision is called. The old behaviour is kept for compatibility purposes. Once, set_sphere_output_precision is called, pgSphere starts to use the old behaviour (read, please, issue #118 discussion). The patch introduces a new function - reset_sphere_output_precision. It is used to reset to the PostgreSQL behaviour after using set_sphere_output_precision. * Update upgrade script (reset_sphere_output_precision function) * Add test for pgSphere output precision with different settings expected/output_precision.out - PG 10-11 expected/output_precision_1.out - PG 12+ * Add extra_float_digits = 2 for epochprop and bounding_box_gist tests
|
The PR was merged. I will close the issue if no objections. |
|
Closed as completed. |
Dear All,
I was wondering if there is any reason to keep the maximum precision of the printed outputs to DBL_DIG macro (15) instead of DBL_DECIMAL_DIG (17), being that just 15 digits don't allow round-trip conversion.
In my case it causes me trouble because I cannot restore dumps of databases done with pg_dump, since the dumps contain spolys which lost precision during the dump, and pg_sphere does not pass the check of certain polygons which were inserted with 17 digits (from caom2db software) but now have 15.
Regards.
The text was updated successfully, but these errors were encountered: