Receiving data in binary format how is it encoded?

Started by Francisco Figueiredo Jr.over 22 years ago5 messages
#1Francisco Figueiredo Jr.
fxjrlists@yahoo.com.br

Hi all,

while implementing the protocol 3.0 I could get the data in text mode
setting the format to 0. (Thanks all for helping me out with flush message)

Now, I want to test the binary format. While some primitive types are
easy to realize, for example a int4 field will be encoded in 4 bytes,
I don't know how other types, like date, time, timestamp, text will be
encoded in binary format.

I tried to find in the code how it would be encoded, but I end up in
some dinamically called functions to do the output and I don't know
where this functions are defined.

Can you provide me some tips where the functions are or the binary
format used to encode the types?

While writing this mail, I saw the docs about binary types... So, my
assumption about int4 be encoded in 4 bytes may be wrong? I mean, the
return of an int4 value, for example 12345 in text would be the chars
12345, but in binary format would be the bytes of 1 2 3 4 5?? Instead of
a 4 bytes int4 value?

The first 4 bytes of bytea is its length? I think I saw something about
this in the code. So in the example below I would get a first 4 bytes
containing the length (5) of the bytes?

Also, just question:
If I have a field which has binary type and I use the extended protocol
to set all result fields to text, will this field also be sent in text,
regardless its type?

Thanks in advance!

--
Regards,

Francisco Figueiredo Jr.

------
"My grandfather once told me that there are two
kinds of people: those
who work and those who take the credit. He told me
to try to be in the
first group; there was less competition there."
- Indira Gandhi

#2Carlos Guzman Alvarez
carlosga@telefonica.net
In reply to: Francisco Figueiredo Jr. (#1)
Re: Receiving data in binary format how is it encoded?

Hello:

while implementing the protocol 3.0 I could get the data in text mode
setting the format to 0. (Thanks all for helping me out with flush message)

Now, I want to test the binary format. While some primitive types are
easy to realize, for example a int4 field will be encoded in 4 bytes,
I don't know how other types, like date, time, timestamp, text will be
encoded in binary format.

Huummmmm i'm going to comment how i make in my C# implementation (stream
is a binary reader ) for example for int4, date and time.

For int4:

int val = stream.ReadInt32();
int int4Value = IPAddress.HostToNetworkOrder(val);

For date ( not sure if i'm making this correct way but values seems to
be correct :) )

DateTime base = new DateTime(2000, 1, 1)
int val = stream.ReadInt32();
int days = IPAddress.HostToNetworkOrder(val);

DateTime date = base.AddDays(days);

For Time:

The same as for date but reading a double and adding it to base as
seconds instead of days.

The only basic data type i'm not handling now is numeric that seems to
be encoded in 10 bytes but i have no seen the postgre source code yet.

--
Best regards

Carlos Guzm�n �lvarez
Vigo-Spain

P.S.: Has you receive an email i sent to you abot little things i see on
npgsql ?? :)

#3Francisco Figueiredo Jr.
fxjrlists@yahoo.com.br
In reply to: Carlos Guzman Alvarez (#2)
Re: Receiving data in binary format how is it encoded?

Carlos Guzman Alvarez wrote:

Hello:

while implementing the protocol 3.0 I could get the data in text mode
setting the format to 0. (Thanks all for helping me out with flush
message)

Now, I want to test the binary format. While some primitive types are
easy to realize, for example a int4 field will be encoded in 4 bytes,
I don't know how other types, like date, time, timestamp, text will be
encoded in binary format.

Huummmmm i'm going to comment how i make in my C# implementation (stream
is a binary reader ) for example for int4, date and time.

Thanks.

For int4:

int val = stream.ReadInt32();
int int4Value = IPAddress.HostToNetworkOrder(val);

Yeah, while waiting for the response, I added support for the int4
datatype. It is really encoded as a 4 bytes value.

For date ( not sure if i'm making this correct way but values seems to
be correct :) )

DateTime base = new DateTime(2000, 1, 1)
int val = stream.ReadInt32();
int days = IPAddress.HostToNetworkOrder(val);

DateTime date = base.AddDays(days);

Uhmmm, where did you take the 2000,1,1 as base from?
If the date is before 2000, days will have a negative value, so date is
calculated correctly backwards?

For Time:

The same as for date but reading a double and adding it to base as
seconds instead of days.

The only basic data type i'm not handling now is numeric that seems to
be encoded in 10 bytes but i have no seen the postgre source code yet.

Thanks Carlos. do you know in what file are these values produced?

--
Regards,

Francisco Figueiredo Jr.

------
"My grandfather once told me that there are two
kinds of people: those
who work and those who take the credit. He told me
to try to be in the
first group; there was less competition there."
- Indira Gandhi

#4Carlos Guzman Alvarez
carlosga@telefonica.net
In reply to: Francisco Figueiredo Jr. (#3)
Re: Receiving data in binary format how is it encoded?

Hello:

Yeah, while waiting for the response, I added support for the int4
datatype. It is really encoded as a 4 bytes value.

:)

Uhmmm, where did you take the 2000,1,1 as base from?
If the date is before 2000, days will have a negative value, so date is
calculated correctly backwards?

In the little tests i do, yes.

Thanks Carlos. do you know in what file are these values produced?

Not sure but i'm reviewing files in directory:

src\backend\utils\adt

This is what seems to be sent for numeric datatype i'm going to test it
now :D:

pq_sendint(&buf, x.ndigits, sizeof(int16));
pq_sendint(&buf, x.weight, sizeof(int16));
pq_sendint(&buf, x.sign, sizeof(int16));
pq_sendint(&buf, x.dscale, sizeof(int16));
for (i = 0; i < x.ndigits; i++)
pq_sendint(&buf, x.digits[i], sizeof(NumericDigit));

--
Best regards

Carlos Guzm�n �lvarez
Vigo-Spain

#5Francisco Figueiredo Jr.
fxjrlists@yahoo.com.br
In reply to: Carlos Guzman Alvarez (#4)
Re: Receiving data in binary format how is it encoded?

Carlos Guzman Alvarez wrote:

Hello:

Hello

Yeah, while waiting for the response, I added support for the int4
datatype. It is really encoded as a 4 bytes value.

:)

Uhmmm, where did you take the 2000,1,1 as base from?
If the date is before 2000, days will have a negative value, so date
is calculated correctly backwards?

In the little tests i do, yes.

Good. I think I found where you get 200,1,1 from:
In datetime.c in the AST dir, there is a line which says:
Assert(POSTGRES_EPOCH_JDATE == date2j(2000, 1, 1));

Thanks Carlos. do you know in what file are these values produced?

Not sure but i'm reviewing files in directory:

src\backend\utils\adt

This is what seems to be sent for numeric datatype i'm going to test it
now :D:

pq_sendint(&buf, x.ndigits, sizeof(int16));
pq_sendint(&buf, x.weight, sizeof(int16));
pq_sendint(&buf, x.sign, sizeof(int16));
pq_sendint(&buf, x.dscale, sizeof(int16));
for (i = 0; i < x.ndigits; i++)
pq_sendint(&buf, x.digits[i], sizeof(NumericDigit));

Thanks. I will have a look in the types in the AST dir.

--
Regards,

Francisco Figueiredo Jr.

------
"My grandfather once told me that there are two
kinds of people: those
who work and those who take the credit. He told me
to try to be in the
first group; there was less competition there."
- Indira Gandhi