Boolean columns - developers-list - monetdb.org

newer
Re: MonetDB: embedded-java - Fix...

Boolean columns

older
Concurrent Transaction control in...

Dimitar Nedev

21 Oct 2015 21 Oct '15

1:07 p.m.

Hey guys, A boolean type column seems to fit in a byte, I guess the extra space is to represent nulls (since it ain't a qubit). But does that mean that no column type is actually ':bit'? Given a BAT, how to distinguish between a byte (tinyint) column and a boolean one? Thanks, Dimitar

Attachments:

signature.asc (application/pgp-signature — 842 bytes)

Reply

Sign in to reply online Use email software

Show replies by date

Stefan Manegold

21 Oct 21 Oct

1:14 p.m.

See https://www.monetdb.org/wiki/MonetDB_type_system ;-) our MAL representation of SQL type BOOLEAN is MAL type :bit, and the C representation of MAL type :bit is (out) C type bit, which is indeed 1 byte (8 bit) wide, because we need to accommodate 3 values: TRUE, FALSE & NULL (NIL). Thus, BOOLEAN columns are (should be) represented by a BAT with tail type (:)bit, which is (semantically) not the same as type TINYINT / (:)bte. Best, Stefan ----- On Oct 21, 2015, at 3:07 PM, Dimitar Nedev dimitar.nedev@monetdbsolutions.com wrote:

Hey guys,

A boolean type column seems to fit in a byte, I guess the extra space is to represent nulls (since it ain't a qubit). But does that mean that no column type is actually ':bit'? Given a BAT, how to distinguish between a byte (tinyint) column and a boolean one?

Thanks, Dimitar

_______________________________________________ developers-list mailing list developers-list@monetdb.org https://www.monetdb.org/mailman/listinfo/developers-list

-- | Stefan.Manegold@CWI.nl | DB Architectures (DA) | | www.CWI.nl/~manegold/ | Science Park 123 (L321) | | +31 (0)20 592-4212 | 1098 XG Amsterdam (NL) |

Reply

Sign in to reply online Use email software

Dimitar Nedev

1:41 p.m.

Thanks for sending it, Stefan. Indeed, the page was quite helpful. That's how I fixed some real/float/double issues earlier. I see, so if I've got a boolean column, calling ATOMstorage(getColumnType(b->T->type)) where 'b' is a BAT would not sufficient to identify it as a boolean? It will return TYPE_bte (2). OK, I'll need to push the logic a bit further. Best, Dimitar

On 2015-Oct-21, at 15:14, Stefan Manegold wrote:

See https://www.monetdb.org/wiki/MonetDB_type_system ;-)

our MAL representation of SQL type BOOLEAN is MAL type :bit, and the C representation of MAL type :bit is (out) C type bit, which is indeed 1 byte (8 bit) wide, because we need to accommodate 3 values: TRUE, FALSE & NULL (NIL).

Thus, BOOLEAN columns are (should be) represented by a BAT with tail type (:)bit, which is (semantically) not the same as type TINYINT / (:)bte.

Best, Stefan

----- On Oct 21, 2015, at 3:07 PM, Dimitar Nedev dimitar.nedev@monetdbsolutions.com wrote:

...
Hey guys,

A boolean type column seems to fit in a byte, I guess the extra space is to represent nulls (since it ain't a qubit). But does that mean that no column type is actually ':bit'? Given a BAT, how to distinguish between a byte (tinyint) column and a boolean one?

Thanks, Dimitar

_______________________________________________ developers-list mailing list developers-list@monetdb.org https://www.monetdb.org/mailman/listinfo/developers-list

-- | Stefan.Manegold@CWI.nl | DB Architectures (DA) | | www.CWI.nl/~manegold/ | Science Park 123 (L321) | | +31 (0)20 592-4212 | 1098 XG Amsterdam (NL) | _______________________________________________ developers-list mailing list developers-list@monetdb.org https://www.monetdb.org/mailman/listinfo/developers-list

Reply

Sign in to reply online Use email software

Sjoerd Mullender

1:50 p.m.

On 21/10/15 15:41, Dimitar Nedev wrote:

Thanks for sending it, Stefan. Indeed, the page was quite helpful. That's how I fixed some real/float/double issues earlier.

I see, so if I've got a boolean column, calling ATOMstorage(getColumnType(b->T->type)) where 'b' is a BAT would not sufficient to identify it as a boolean? It will return TYPE_bte (2).

The name ATOMstorage should tell you a lot: it's about *storage*, not semantics. A bit uses the same amount of storage as a bte, but that doesn't mean those bytes are interpreted the same. It may well be that you shouldn't use ATOMstorage here.

OK, I'll need to push the logic a bit further.

Best, Dimitar

...
On 2015-Oct-21, at 15:14, Stefan Manegold wrote:

See https://www.monetdb.org/wiki/MonetDB_type_system ;-)

our MAL representation of SQL type BOOLEAN is MAL type :bit, and the C representation of MAL type :bit is (out) C type bit, which is indeed 1 byte (8 bit) wide, because we need to accommodate 3 values: TRUE, FALSE & NULL (NIL).

Thus, BOOLEAN columns are (should be) represented by a BAT with tail type (:)bit, which is (semantically) not the same as type TINYINT / (:)bte.

Best, Stefan

----- On Oct 21, 2015, at 3:07 PM, Dimitar Nedev dimitar.nedev@monetdbsolutions.com wrote:

...
Hey guys,

A boolean type column seems to fit in a byte, I guess the extra space is to represent nulls (since it ain't a qubit). But does that mean that no column type is actually ':bit'? Given a BAT, how to distinguish between a byte (tinyint) column and a boolean one?

Thanks, Dimitar

_______________________________________________ developers-list mailing list developers-list@monetdb.org https://www.monetdb.org/mailman/listinfo/developers-list

-- | Stefan.Manegold@CWI.nl | DB Architectures (DA) | | www.CWI.nl/~manegold/ | Science Park 123 (L321) | | +31 (0)20 592-4212 | 1098 XG Amsterdam (NL) | _______________________________________________ developers-list mailing list developers-list@monetdb.org https://www.monetdb.org/mailman/listinfo/developers-list

_______________________________________________ developers-list mailing list developers-list@monetdb.org https://www.monetdb.org/mailman/listinfo/developers-list

-- Sjoerd Mullender

Reply

Sign in to reply online Use email software

Stefan Manegold

1:54 p.m.

of a type t, ATOMstorage(t) gives you the "normalized" type with the same storage size; getColumnType() gives you the actual type. ATOMstorage(t) is usually used to reduce the type-specific code expansion, if only the size matters (e.g., for copying or bit-wise equality comparison), rather than the full type semantics ----- On Oct 21, 2015, at 3:41 PM, Dimitar Nedev dimitar.nedev@monetdbsolutions.com wrote:

Thanks for sending it, Stefan. Indeed, the page was quite helpful. That's how I fixed some real/float/double issues earlier.

I see, so if I've got a boolean column, calling ATOMstorage(getColumnType(b->T->type)) where 'b' is a BAT would not sufficient to identify it as a boolean? It will return TYPE_bte (2).

OK, I'll need to push the logic a bit further.

Best, Dimitar

...
On 2015-Oct-21, at 15:14, Stefan Manegold wrote:

See https://www.monetdb.org/wiki/MonetDB_type_system ;-)

our MAL representation of SQL type BOOLEAN is MAL type :bit, and the C representation of MAL type :bit is (out) C type bit, which is indeed 1 byte (8 bit) wide, because we need to accommodate 3 values: TRUE, FALSE & NULL (NIL).

Thus, BOOLEAN columns are (should be) represented by a BAT with tail type (:)bit, which is (semantically) not the same as type TINYINT / (:)bte.

Best, Stefan

----- On Oct 21, 2015, at 3:07 PM, Dimitar Nedev dimitar.nedev@monetdbsolutions.com wrote:

...
Hey guys,

A boolean type column seems to fit in a byte, I guess the extra space is to represent nulls (since it ain't a qubit). But does that mean that no column type is actually ':bit'? Given a BAT, how to distinguish between a byte (tinyint) column and a boolean one?

Thanks, Dimitar

_______________________________________________ developers-list mailing list developers-list@monetdb.org https://www.monetdb.org/mailman/listinfo/developers-list

-- | Stefan.Manegold@CWI.nl | DB Architectures (DA) | | www.CWI.nl/~manegold/ | Science Park 123 (L321) | | +31 (0)20 592-4212 | 1098 XG Amsterdam (NL) | _______________________________________________ developers-list mailing list developers-list@monetdb.org https://www.monetdb.org/mailman/listinfo/developers-list

_______________________________________________ developers-list mailing list developers-list@monetdb.org https://www.monetdb.org/mailman/listinfo/developers-list

-- | Stefan.Manegold@CWI.nl | DB Architectures (DA) | | www.CWI.nl/~manegold/ | Science Park 123 (L321) | | +31 (0)20 592-4212 | 1098 XG Amsterdam (NL) |

Reply

Sign in to reply online Use email software

3304

Age (days ago)

3304

Last active (days ago)

Download

4 comments

3 participants

tags

participants (3)

Dimitar Nedev
Sjoerd Mullender
Stefan Manegold