Configurable floats format on conversions #130
Conversation
- in case password generation fails, log the messages generated by the script. - also, in case another Elasticsearch instance is running, fail with the correct error message.
- in case there's an unexpected data type received from Elasticsearch (ex.: float value for DATE) make sure the application receives a specific diagnostic.
This commit adds a new config paramter to the connection string that instructs the driver to convert the floats into the default, scientific or auto-mode notation; auto meaning scientific if the value is small or large "enough". It also uses scale received from Elasticsearch for the respective type. Also, on the consistency checks the driver does on descriptors, the precision and scale are now also those read from Elasticsearch (rather than local values).
driver/convert.c
Outdated
| static SQLRETURN double_to_str(esodbc_rec_st *arec, esodbc_rec_st *irec, | ||
| double dbl, void *data_ptr, SQLLEN *octet_len_ptr, BOOL wide) | ||
| { | ||
| # define DBL_BASE10_MAX_LEN /*-0.*/3 + DBL_MANT_DIG - DBL_MIN_EXP |
There was a problem hiding this comment.
I believe DBL_MANT_DIG and DBL_MIN_EXP refer to the base 2 representation, not base 10, so they're 53 and -1021 respectively. The base 10 equivalents are 15 and -308. In other words the number of characters to print the smallest double in fixed point notation would be 3 + 15 - -308, not 3 + 53 - -1021. I think the correct constants in C99 are DBL_DIG and DBL_MIN_10_EXP. But if Microsoft's headers don't define these or if you want to be allow a bit more in case there's some edge case where more digits are printed then maybe just define a constant to slightly higher like 336 (= 21 * 16) and use that for the buffer size including the \0 terminator?
There was a problem hiding this comment.
Thanks for spotting that, fixed now.
The defs are available (https://docs.microsoft.com/en-us/cpp/cpp/floating-limits).
I've stuck with the "exact" calculation, since the buffer should anyways be way over-provisioned for the precision (=ODBC's scale) these numbers will be printed with. (Currently the limit is 19, but even that is probably too high -- pending further clarifications.)
- addressing PR comments.
* integ. test.: more verbosity on pwd gen. failure - in case password generation fails, log the messages generated by the script. - also, in case another Elasticsearch instance is running, fail with the correct error message. * fix: return error upstream in case of API errors - in case there's an unexpected data type received from Elasticsearch (ex.: float value for DATE) make sure the application receives a specific diagnostic. * add configurable mode for floats conversion This commit adds a new config paramter to the connection string that instructs the driver to convert the floats into the default, scientific or auto-mode notation; auto meaning scientific if the value is small or large "enough". It also uses scale received from Elasticsearch for the respective type. Also, on the consistency checks the driver does on descriptors, the precision and scale are now also those read from Elasticsearch (rather than local values). * fix: use radix 10 for dbl buff limits - addressing PR comments. (cherry picked from commit d61f3cf)
bpintea
added
>feature
This PR adds a new config paramter to the connection string that instructs the driver to convert the floats into the default, scientific or auto-mode notation; auto meaning scientific if the value is small or
large "enough".
It also uses scale received from Elasticsearch for the respective type.
A defect was corrected here, the application now no longer being able to specify the scale for conversion (this is source specific).
Also, on the consistency checks the driver does on descriptors, the precision and scale are now also those read from Elasticsearch (rather than local values).