Bit of an obscure bug this one. SQLite allows column names to contain [ and ] characters, even though those are often used as delimiters in SQLite. Here's how to create such a database with bash:```sqlite3 /tmp/demo.db <<EOF    BEGIN TRANSACTION;    CREATE TABLE "data" (        "MTU (CET)" TEXT,        "Day-ahead Price [EUR/MWh]" TEXT    );    INSERT INTO "data" VALUES('01.01.2016 00:00 - 01.01.2016 01:00','23.86');    COMMIT;EOF```If you query that database from sqlite3 in Python, the `[Eur/MWh]` piece is removed from that column name:```In [1]: import sqlite3                                                                                                         In [2]: conn = sqlite3.connect("/tmp/demo.db")                                                                                 In [3]: cursor = conn.cursor()                                                                                                 In [4]: cursor.execute("select * from data")                                                                                   Out[4]: <sqlite3.Cursor at 0x10c70a0a0>In [5]: cursor.fetchall()                                                                                                      Out[5]: [('01.01.2016 00:00 - 01.01.2016 01:00', '23.86')]In [6]: cursor.description                                                                                                     Out[6]: (('MTU (CET)', None, None, None, None, None, None), ('Day-ahead Price', None, None, None, None, None, None))In [7]: conn.row_factory = sqlite3.Row                                                                                         In [8]: cursor = conn.cursor()                                                                                                 In [9]: cursor.execute("select * from data")                                                                                   Out[9]: <sqlite3.Cursor at 0x10c7a8490>In [10]: row = cursor.fetchall()                                                                                                     In [12]: row                                                                                                                   Out[12]: <sqlite3.Row at 0x10c3fe670>In [15]: row.keys()                                                                                                            Out[15]: ['MTU (CET)', 'Day-ahead Price']```As you can see, it is missing from both `cursor.description` and from `row.keys()` here.But... if you query that database using SQLite directly (with `.headers on` so you can see the name of the columns) it works as expected:```$ sqlite3 /tmp/demo.db SQLite version 3.24.0 2018-06-04 14:10:15Enter ".help" for usage hints.sqlite> .schemaCREATE TABLE IF NOT EXISTS "data" (        "MTU (CET)" TEXT,        "Day-ahead Price [EUR/MWh]" TEXT    );sqlite> .headers onsqlite> select * from data;MTU (CET)|Day-ahead Price [EUR/MWh]01.01.2016 00:00 - 01.01.2016 01:00|23.86sqlite> ```It looks to me like this is a bug in Python's SQLite3 module.This was first reported here:https://github.com/simonw/sqlite-utils/issues/86

Thanks for reporting this issue, I have a PR herehttps://github.com/python/cpython/pull/18533 but that still needs some clarification, because there were some tests that were actually testing that we were stripping everything after the square brackets.P.D. this is my first contribution :)

Oh how interesting - yes it looks like this is deliberate behavior introduced in this commit:https://github.com/python/cpython/commit/0e3f591aeeef9ed715f8770320f4c4c7332a8794

https://github.com/ghaering/pysqlite/commit/f3d452f2daeb432b8ad89fa4f087164bfd6ddc12 should probably give more context than that huge svnmerge commit :)(Removed older Python versions since even if decide to change the behavior from 2006, it may break third-party programs that are relying on it.)

It was originally introduced inhttps://github.com/ghaering/pysqlite/commit/780a76dcd8f4142b6ecbd423f52e0ccf067fc277. I think that original column names should be returned when PARSE_COLNAMES is not set.Working on a PR.

hi @serhiy.storchaka is this something that you think could be done by a new contributor? I'd really love to take care of this, I can improve on the PR I was preparinghttps://github.com/python/cpython/pull/18533

Sorry Raul, but I already have a PR written. It includes minor refactoring,  better error handling in the nearby code, improved tests, fixed documentation and comments which incorrectly described the current behavior. It took time to determine what parts should be fixed, and unless you have found all this in your PR it would take much more time to explain to you what details should be checked than check and fix them myself.

New changesetb146568dfcbcd7409c724f8917e4f77433dd56e4 by Serhiy Storchaka in branch 'master':bpo-39652: Truncate the column name after '[' only if PARSE_COLNAMES is set. (GH-18942)https://github.com/python/cpython/commit/b146568dfcbcd7409c724f8917e4f77433dd56e4

New changeset687f5921a46cf95c2a648d8031f9e99cdcc3e6b7 by Miss Islington (bot) in branch '3.8':bpo-39652: Truncate the column name after '[' only if PARSE_COLNAMES is set. (GH-18942)https://github.com/python/cpython/commit/687f5921a46cf95c2a648d8031f9e99cdcc3e6b7

New changeset39680fb7043e555469e08d3c4f49073acca77b20 by Serhiy Storchaka in branch '3.7':[3.7]bpo-39652: Truncate the column name after '[' only if PARSE_COLNAMES is set. (GH-18942). (GH-19104)https://github.com/python/cpython/commit/39680fb7043e555469e08d3c4f49073acca77b20