Verified our report against the Комита reference export by joining on the
record id. Our report_20260923_5.xlsx covers the 114 files of 30.01.2023 and
all 5266 of its records are inside the reference period, so the two can be
compared row by row.
Fifteen of the twenty-one shared columns already match exactly. The largest
remaining gap was the separator in the code lists: the reference prints
"1101,1106,1499" while the parser joins repeated tags with "; ", which put
336 of 5266 records out of step. Only the two code-list columns are brought
to the reference's separator; "; " stays in the free-text columns, where it
reads better and where the reference has no equivalent.
What the comparison also showed, for the record:
- The empty branch БИК and number on head-office records are correct. Those
files carry no ИнфФилиал block at all and the reference leaves the cells
empty too.
- Of the 105 address differences, 94 are the column 141 leak fixed in
cceee02, and the remaining 11 are values that contain a comma, such as
Пункт = "Южно-Сахалинск г, Хомутово пл.р-н", which the reference cannot
represent in one comma-joined cell.
- Column 121 shows РегнНомер for people in private practice, where the
reference's ОГРН column stays empty. That is the mapping table's fourth
source for the column, not a defect: 3 records out of 13168.
69fd329f78