Help with Perl While Loop
Is there any way to make this script run faster? Right now it is taking almost an hour to
process a 91,000 flat file. Here is the while loop code which runs through each line in the
file and insert it into an oracle database table.
my $line = 0;
my $c = 0;
while (<IN>) {
chomp($_);
$line += 1;
my ($field1,
$field2,
$field3,
$field4,
$field5,
$field6,
$field7,
$field8,
$field9,
$field10,
$field11,
$field12,
$field13,
$field14,
$field15,
$field16,
$field17,
$field18,
$field19,
$field20,
$field21,
$field22,
$field23,
$field24) = split(/ /,$_);
&fix($field1);
&fix($field2);
&fix($field3);
&fix($field4);
&fix($field5);
&fix($field6);
&fix($field7);
&fix($field8);
&fix($field9);
&fix($field10);
&fix($field11);
&fix($field12);
&fix($field13);
&fix($field14);
&fix($field15);
&fix($field16);
&fix($field17);
&fix($field18);
&fix($field19);
&fix($field20);
&fix($field21);
&fix($field22);
&fix($field23);
&fix($field24);
if ($field1 ne ' ') {
$field1 =~ s/\///g;
my ($month, $day, $year) = unpack "A2A2A4", $field1;
$field1 = $year.''.$month.''.$day;
}
$field23 = lc($field23);
$field24 = lc($field24);
if (($field15 ne "A") && ($field16 ne "B")) {
eval {
$sql = "insert into myTable values (\'$field1\',\'$field2\',\
'$field8\',\'$field9\',\'$
'$field17\',\'$field18\',\
$sth = $dbh->prepare($sql);
$sth->execute();
};
if ($@) { print "$@\n"; next; }
$dbh->commit();
$c += 1;
}
}
Subroutine that's called -
sub fix {
$_[0] =~ s/^ *//;
$_[0] =~ s/ *$//;
$_[0] =~ s/'/''/g;
if (! $_[0]) {
$_[0] = ' ';
}
}
process a 91,000 flat file. Here is the while loop code which runs through each line in the
file and insert it into an oracle database table.
my $line = 0;
my $c = 0;
while (<IN>) {
chomp($_);
$line += 1;
my ($field1,
$field2,
$field3,
$field4,
$field5,
$field6,
$field7,
$field8,
$field9,
$field10,
$field11,
$field12,
$field13,
$field14,
$field15,
$field16,
$field17,
$field18,
$field19,
$field20,
$field21,
$field22,
$field23,
$field24) = split(/ /,$_);
&fix($field1);
&fix($field2);
&fix($field3);
&fix($field4);
&fix($field5);
&fix($field6);
&fix($field7);
&fix($field8);
&fix($field9);
&fix($field10);
&fix($field11);
&fix($field12);
&fix($field13);
&fix($field14);
&fix($field15);
&fix($field16);
&fix($field17);
&fix($field18);
&fix($field19);
&fix($field20);
&fix($field21);
&fix($field22);
&fix($field23);
&fix($field24);
if ($field1 ne ' ') {
$field1 =~ s/\///g;
my ($month, $day, $year) = unpack "A2A2A4", $field1;
$field1 = $year.''.$month.''.$day;
}
$field23 = lc($field23);
$field24 = lc($field24);
if (($field15 ne "A") && ($field16 ne "B")) {
eval {
$sql = "insert into myTable values (\'$field1\',\'$field2\',\
'$field8\',\'$field9\',\'$
'$field17\',\'$field18\',\
$sth = $dbh->prepare($sql);
$sth->execute();
};
if ($@) { print "$@\n"; next; }
$dbh->commit();
$c += 1;
}
}
Subroutine that's called -
sub fix {
$_[0] =~ s/^ *//;
$_[0] =~ s/ *$//;
$_[0] =~ s/'/''/g;
if (! $_[0]) {
$_[0] = ' ';
}
}
The perl code can be streamlined, but I would expect that most of the time would be taken by the $sth->execute();
You might try moving the prepare out of the loop and doing a bind in the loop, or only doing a commit after several inserts.
You might try moving the prepare out of the loop and doing a bind in the loop, or only doing a commit after several inserts.
There are a number of inefficiencies and questionable coding practices in that code, but without knowing where the script is spending most of its time, we can't say for certain which of those inefficiencies you should focus on fixing.
You can and should use the Devel::NYTProf module to learn where the script is spending most of its time. We can also make some educated guesses, like ozo did, and correct some of those inefficiencies.
If you use ' ' for the pattern in the split statement instead of / /, it will strip the leading and trailing spaces from the fields, so you won't need to do that in the fix() sub. And, instead of creating 24 sequentially numbered scalars, it would be better to use an array.
Instead of calling the fix() sub 24 times for each line in the file, it would be better to alter the sub to have it accept an array reference.
ozo has already suggested moving the prepare statement, which should defiantly be done. When doing that, you need to use placeholders in the prepare statement instead of passing the $fieldX vars. Those vars (or the array) would be passed in the execute statement.
Placeholders and Bind Values
There are additional improvements that I can suggest, but start with those and we'll cover the others as needed.
You can and should use the Devel::NYTProf module to learn where the script is spending most of its time. We can also make some educated guesses, like ozo did, and correct some of those inefficiencies.
If you use ' ' for the pattern in the split statement instead of / /, it will strip the leading and trailing spaces from the fields, so you won't need to do that in the fix() sub. And, instead of creating 24 sequentially numbered scalars, it would be better to use an array.
Instead of calling the fix() sub 24 times for each line in the file, it would be better to alter the sub to have it accept an array reference.
my @fields = split(' ', $_);
fix(\@fields);
ozo has already suggested moving the prepare statement, which should defiantly be done. When doing that, you need to use placeholders in the prepare statement instead of passing the $fieldX vars. Those vars (or the array) would be passed in the execute statement.
Placeholders and Bind Values
There are additional improvements that I can suggest, but start with those and we'll cover the others as needed.
my $c = 0;
my @field;
$sth = $dbh->prepare("insert into myTable values (?,?,?,?,?,?,?,?,?,?,?,?.?
while (<IN>) {
@field[1..24] = split;
next unless $field[15] ne "A" && $field[16] ne "B";
tr/'/"/, $_||=' ' for @field;
$_=join'',(split m{/*})[4..7,0..3] for $field[1];
$_ = lc for @field[23,24];
eval {
$sth->execute(@field[1..24
};
if ($@) { print "$@\n"; next; }
$dbh->commit();
$c += 1;
}
my $line = $.;
my @field;
$sth = $dbh->prepare("insert into myTable values (?,?,?,?,?,?,?,?,?,?,?,?.?
while (<IN>) {
@field[1..24] = split;
next unless $field[15] ne "A" && $field[16] ne "B";
tr/'/"/, $_||=' ' for @field;
$_=join'',(split m{/*})[4..7,0..3] for $field[1];
$_ = lc for @field[23,24];
eval {
$sth->execute(@field[1..24
};
if ($@) { print "$@\n"; next; }
$dbh->commit();
$c += 1;
}
my $line = $.;
Do you need Perl at all?
When loading data from a text file into Oracle SQL Loader or External Tables is typically the best/fastest.
When loading data from a text file into Oracle SQL Loader or External Tables is typically the best/fastest.
I think so. I need to process the file on a daily basis out of cron.
I'm having problems with the split splitting properly as some fields have spaces between their values.
Example of a line in the file:
01/01/2013 123456789 username01 P abc@email.com Alfa B Canine Alfa B Canine 123 South 1st Street New York NY 10000 Some Park Main +1 212 555 1234 +1 212 555 4321 75 Some Blvd New York NY 10000 Some Really Long Org Name Some Really Long Org Name ABC+D Water Gone Status_Range
I'm trying -
my @field;
$sth = $dbh->prepare("insert into myTable values (?,?,?,?,?,?,?,?,?,?,?,?,?
while (<IN>) {
@field[1..37] = split(/ /, $_);
next unless $field[29] ne "Some Really Long Org Name" && $field[30] ne "Some Really Long Org Name";
tr/'/"/, $_||=' ' for @field;
$_=join'',(split m{/*})[4..7,0..3] for $field[1];
$_ = lc for @field[32,33];
eval {
$sth->execute(@field[1..37
};
if ($@) { print "$@\n"; next; }
$dbh->commit();
}
There are actually 37 fields in this file not 24. :) Also, for the second field (123456789, in this example) I'd like to substitute "None" (without the quotes) if the field is blank.
I'm having problems with the split splitting properly as some fields have spaces between their values.
Example of a line in the file:
01/01/2013 123456789 username01 P abc@email.com Alfa B Canine Alfa B Canine 123 South 1st Street New York NY 10000 Some Park Main +1 212 555 1234 +1 212 555 4321 75 Some Blvd New York NY 10000 Some Really Long Org Name Some Really Long Org Name ABC+D Water Gone Status_Range
I'm trying -
my @field;
$sth = $dbh->prepare("insert into myTable values (?,?,?,?,?,?,?,?,?,?,?,?,?
while (<IN>) {
@field[1..37] = split(/ /, $_);
next unless $field[29] ne "Some Really Long Org Name" && $field[30] ne "Some Really Long Org Name";
tr/'/"/, $_||=' ' for @field;
$_=join'',(split m{/*})[4..7,0..3] for $field[1];
$_ = lc for @field[32,33];
eval {
$sth->execute(@field[1..37
};
if ($@) { print "$@\n"; next; }
$dbh->commit();
}
There are actually 37 fields in this file not 24. :) Also, for the second field (123456789, in this example) I'd like to substitute "None" (without the quotes) if the field is blank.
>>I'd like to substitute "None" (without the quotes) if the field is blank.
Databases allow null values for a reason. Use them. Why store data to say no data exists?
Databases allow null values for a reason. Use them. Why store data to say no data exists?
True, true. I'm in favor of that! Unfortunately, this is legacy table (which does not allow nulls -ech), and legacy code that I am trying to improve in small steps.
I hate bad designs that you are forced to live with.
Anyway, did you look into SQL Loader to replace the Perl Script? SQL Loader was written to load data into Oracle from Text files and does this very efficiently.
Anyway, did you look into SQL Loader to replace the Perl Script? SQL Loader was written to load data into Oracle from Text files and does this very efficiently.
> some fields have spaces between their values.
Your use of / / instead of ' ' suggested that you wanted multiple spaces to indicate empty fields.
If fields can be separated by multiple spaces, you probably want to use " ", which is the default:
@field[1..37] = split;
Why wouldn't you be able to use SQL Loader with cron?
Your use of / / instead of ' ' suggested that you wanted multiple spaces to indicate empty fields.
If fields can be separated by multiple spaces, you probably want to use " ", which is the default:
@field[1..37] = split;
Why wouldn't you be able to use SQL Loader with cron?
> if the field is blank.
If there can be multiple spaces between fields, how do you represent a blank field?
If the number of spaces between fields is always between 4 and 8,
so less than 4 spaces in a row are part of a field, and 19 spaces in a row, as between Canine and 123 indicates an empty field, then you might use
@field[1..37] = split/ {4,8}/;
If there can be multiple spaces between fields, how do you represent a blank field?
If the number of spaces between fields is always between 4 and 8,
so less than 4 spaces in a row are part of a field, and 19 spaces in a row, as between Canine and 123 indicates an empty field, then you might use
@field[1..37] = split/ {4,8}/;
sqlldr is only command line.
Setting up the control file is the hard part.
Once you have it set up the actual command is simple:
sqlldr username/password control=mycontrolfile.ctl
You've provided an example line from the file. If you can provide the table description and the expected results I'm sure we can come up with a sample control file.
Setting up the control file is the hard part.
Once you have it set up the actual command is simple:
sqlldr username/password control=mycontrolfile.ctl
You've provided an example line from the file. If you can provide the table description and the expected results I'm sure we can come up with a sample control file.
With split/ / or split' ', the s/^ *// and s/ *$// would have been unnecessary, since the spaces would separate the fields instead of being part of the field, but with something like split/ {4,8}/, the s/^ *// may become necessary
s/^ *//, tr/'/"/, $_||=' ' for @field;
And it may have to be done before the "Some Really Long Org Name" check,
or, depending on how often that case happens, it may be faster to make the check allow for leading space:
next if grep/ *Some Really Long Org Name$/,@field[29,30];
s/^ *//, tr/'/"/, $_||=' ' for @field;
And it may have to be done before the "Some Really Long Org Name" check,
or, depending on how often that case happens, it may be faster to make the check allow for leading space:
next if grep/ *Some Really Long Org Name$/,@field[29,30];
>Is there any way to not have to pass the user/password from the command line?
Looks like you can store it in a PARFILE as well. Never used PARFILEs with sql loader but the docs say you can.
Looks like you can store it in a PARFILE as well. Never used PARFILEs with sql loader but the docs say you can.
Hmm.. I think I'll work on the sql loader way after I get this ancient way with a Perl script working. :)
Personally I would stop the ancient way and focus efforts on SQL Loader. Trust me, it is pretty quick at what it does.
If you can: Create a very simple test case. Single table, single column. Should be a simple tweak to your Perl script to handle it and a very simple sql loader script.
Set up some representative sample of rows that you expect in production and test them.
If you can: Create a very simple test case. Single table, single column. Should be a simple tweak to your Perl script to handle it and a very simple sql loader script.
Set up some representative sample of rows that you expect in production and test them.
Ok, will tackle the SQL Loader way! :)
Here is the table description:
Name Null Type
--------------- -------- ------------
FIELD1 NOT NULL VARCHAR2(8) # date of birth in file is mm/dd/yyyy but need to reformat to yyyymmdd
FIELD2 NOT NULL VARCHAR2(9) # replace blanks with None
FIELD3 NOT NULL VARCHAR2(16)
FIELD4 NOT NULL VARCHAR2(2)
FIELD5 NOT NULL VARCHAR2(64)
FIELD6 NOT NULL VARCHAR2(31)
FIELD7 NOT NULL VARCHAR2(20)
FIELD8 NOT NULL VARCHAR2(32)
FIELD9 NOT NULL VARCHAR2(8)
FIELD10 NOT NULL VARCHAR2(16)
FIELD11 NOT NULL VARCHAR2(31)
FIELD12 NOT NULL VARCHAR2(20)
FIELD13 NOT NULL VARCHAR2(32)
FIELD14 NOT NULL VARCHAR2(16)
FIELD15 NOT NULL VARCHAR2(16)
FIELD16 NOT NULL VARCHAR2(96)
FIELD17 NOT NULL VARCHAR2(32)
FIELD18 NOT NULL VARCHAR2(32)
FIELD19 NOT NULL VARCHAR2(16)
FIELD20 NOT NULL VARCHAR2(64)
FIELD21 NOT NULL VARCHAR2(24)
FIELD22 NOT NULL VARCHAR2(24)
FIELD23 NOT NULL VARCHAR2(32)
FIELD24 NOT NULL VARCHAR2(32)
FIELD25 NOT NULL VARCHAR2(96)
FIELD26 NOT NULL VARCHAR2(64)
FIELD27 NOT NULL VARCHAR2(32)
FIELD28 NOT NULL VARCHAR2(16)
FIELD29 NOT NULL VARCHAR2(64) # Some Really Long Org Name
FIELD30 NOT NULL VARCHAR2(64) # Some Really Long Org Name
FIELD31 NOT NULL VARCHAR2(64)
FIELD32 NOT NULL VARCHAR2(16)
FIELD33 NOT NULL VARCHAR2(24)
FIELD34 NOT NULL VARCHAR2(16)
FIELD35 NOT NULL VARCHAR2(1)
FIELD36 NOT NULL VARCHAR2(1)
FIELD37 NOT NULL VARCHAR2(1)
Fields 6-8 are names and sometimes have apostrophes.
Here is the table description:
Name Null Type
--------------- -------- ------------
FIELD1 NOT NULL VARCHAR2(8) # date of birth in file is mm/dd/yyyy but need to reformat to yyyymmdd
FIELD2 NOT NULL VARCHAR2(9) # replace blanks with None
FIELD3 NOT NULL VARCHAR2(16)
FIELD4 NOT NULL VARCHAR2(2)
FIELD5 NOT NULL VARCHAR2(64)
FIELD6 NOT NULL VARCHAR2(31)
FIELD7 NOT NULL VARCHAR2(20)
FIELD8 NOT NULL VARCHAR2(32)
FIELD9 NOT NULL VARCHAR2(8)
FIELD10 NOT NULL VARCHAR2(16)
FIELD11 NOT NULL VARCHAR2(31)
FIELD12 NOT NULL VARCHAR2(20)
FIELD13 NOT NULL VARCHAR2(32)
FIELD14 NOT NULL VARCHAR2(16)
FIELD15 NOT NULL VARCHAR2(16)
FIELD16 NOT NULL VARCHAR2(96)
FIELD17 NOT NULL VARCHAR2(32)
FIELD18 NOT NULL VARCHAR2(32)
FIELD19 NOT NULL VARCHAR2(16)
FIELD20 NOT NULL VARCHAR2(64)
FIELD21 NOT NULL VARCHAR2(24)
FIELD22 NOT NULL VARCHAR2(24)
FIELD23 NOT NULL VARCHAR2(32)
FIELD24 NOT NULL VARCHAR2(32)
FIELD25 NOT NULL VARCHAR2(96)
FIELD26 NOT NULL VARCHAR2(64)
FIELD27 NOT NULL VARCHAR2(32)
FIELD28 NOT NULL VARCHAR2(16)
FIELD29 NOT NULL VARCHAR2(64) # Some Really Long Org Name
FIELD30 NOT NULL VARCHAR2(64) # Some Really Long Org Name
FIELD31 NOT NULL VARCHAR2(64)
FIELD32 NOT NULL VARCHAR2(16)
FIELD33 NOT NULL VARCHAR2(24)
FIELD34 NOT NULL VARCHAR2(16)
FIELD35 NOT NULL VARCHAR2(1)
FIELD36 NOT NULL VARCHAR2(1)
FIELD37 NOT NULL VARCHAR2(1)
Fields 6-8 are names and sometimes have apostrophes.
I looked at the file in vi with :set list and see that the fields are separated with ^I
and lines are terminated with ^M$
I was removing the ctrl-Ms with: perl -pi -e 's/\cM//g' $1
and lines are terminated with ^M$
I was removing the ctrl-Ms with: perl -pi -e 's/\cM//g' $1
>>I looked at the file in vi with :set list and see that the fields are separated with ^I
Tab delimited makes sql loader even easier.
I'm about ready to leave work. I'll work on this later at home. Shouldn't be difficult. Give me a little commute time and I should have something for you in an hour or two.
Can you post some actual tab delimited data (in a code block) for me?
It will cut down on errors when I add them to what you already posted.
Tab delimited makes sql loader even easier.
I'm about ready to leave work. I'll work on this later at home. Shouldn't be difficult. Give me a little commute time and I should have something for you in an hour or two.
Can you post some actual tab delimited data (in a code block) for me?
This is a code block
It will cut down on errors when I add them to what you already posted.
Wooo, thank you!! :)
Here's a line!
Here's a line!
01/01/2013 123456789 username01 P abc@email.com Alfa B Canine Alfa B Canine 123 South 1st Street New York NY
10000 Some Park Main +1 212 555 1234 +1 212 555 4321 75 Some Blvd New York NY 10000 Some Really Long Org Name Some Really Long Org Name ABC+D Water Gone Status_Range Y ^M
>>Here's a line!
This still doesn't have the tab separation. I'll make some guesses. It should be enough to get you off and running.
This still doesn't have the tab separation. I'll make some guesses. It should be enough to get you off and running.
There doesn't appear to be enough fields in the data file for the table definition so its a little off.
Can you also change how the file gets to the Unix box? The ctrl-M is common when files are tranferred from DOS/Windows to Unix. doscp or similar command can help with this.
Assuming you cannot, check out FIELD37 in my control file. The terminated by syntax should work once all the fields line up.
I also added another row based on your requirements for 'None' and possible single quotes in the data.
Here's what I have:
Control File (I named mine q.ctl):
My data file (named q.txt. I added tabs where I thought they should go):
PARFILE (name q.par):
My table setup:
The command from the prompt:
Can you also change how the file gets to the Unix box? The ctrl-M is common when files are tranferred from DOS/Windows to Unix. doscp or similar command can help with this.
Assuming you cannot, check out FIELD37 in my control file. The terminated by syntax should work once all the fields line up.
I also added another row based on your requirements for 'None' and possible single quotes in the data.
Here's what I have:
Control File (I named mine q.ctl):
LOAD DATA
INFILE "c:\q.txt"
APPEND
INTO TABLE tab1
FIELDS TERMINATED BY X'9'
TRAILING NULLCOLS
(
FIELD1 "substr(:FIELD1,7,4) || substr(:FIELD1,1,2) || substr(:FIELD1,4,2)",
FIELD2 "NVL(:FIELD2,'None')",
FIELD3,
FIELD4,
FIELD5,
FIELD6,
FIELD7,
FIELD8,
FIELD9,
FIELD10,
FIELD11,
FIELD12,
FIELD13,
FIELD14,
FIELD15,
FIELD16,
FIELD17,
FIELD18,
FIELD19,
FIELD20,
FIELD21,
FIELD22,
FIELD23,
FIELD24,
FIELD25,
FIELD26,
FIELD27,
FIELD28,
FIELD29,
FIELD30,
FIELD31,
FIELD32,
FIELD33,
FIELD34,
FIELD35,
FIELD36,
FIELD37 TERMINATED BY X'D'
)
My data file (named q.txt. I added tabs where I thought they should go):
01/01/2013 123456789 username01 P abc@email.com Alfa B Canine Alfa B Canine 123 South 1st Street New York NY 10000 Some Park Main +1 212 555 1234 +1 212 555 4321 75 Some Blvd New York NY 10000 Some Really Long Org Name Some Really Long Org Name ABC+D Water Gone Status_Range Y ^M
02/02/2013 mytest P abc@email.com Al'fa B Canine Alfa B Canine 123 South 1st Street New York NY 10000 Some Park Main +1 212 555 1234 +1 212 555 4321 75 Some Blvd New York NY 10000 Some Really Long Org Name Some Really Long Org Name ABC+D Water Gone Status_Range Y ^M
PARFILE (name q.par):
userid=scott/tiger
control=q.ctl
My table setup:
create table tab1(
FIELD1 VARCHAR2(8),
FIELD2 VARCHAR2(9),
FIELD3 VARCHAR2(16),
FIELD4 VARCHAR2(2),
FIELD5 VARCHAR2(64),
FIELD6 VARCHAR2(31),
FIELD7 VARCHAR2(20),
FIELD8 VARCHAR2(32),
FIELD9 VARCHAR2(8),
FIELD10 VARCHAR2(16),
FIELD11 VARCHAR2(31),
FIELD12 VARCHAR2(20),
FIELD13 VARCHAR2(32),
FIELD14 VARCHAR2(16),
FIELD15 VARCHAR2(16),
FIELD16 VARCHAR2(96),
FIELD17 VARCHAR2(32),
FIELD18 VARCHAR2(32),
FIELD19 VARCHAR2(16),
FIELD20 VARCHAR2(64),
FIELD21 VARCHAR2(24),
FIELD22 VARCHAR2(24),
FIELD23 VARCHAR2(32),
FIELD24 VARCHAR2(32),
FIELD25 VARCHAR2(96),
FIELD26 VARCHAR2(64),
FIELD27 VARCHAR2(32),
FIELD28 VARCHAR2(16),
FIELD29 VARCHAR2(64),
FIELD30 VARCHAR2(64),
FIELD31 VARCHAR2(64),
FIELD32 VARCHAR2(16),
FIELD33 VARCHAR2(24),
FIELD34 VARCHAR2(16),
FIELD35 VARCHAR2(1),
FIELD36 VARCHAR2(1),
FIELD37 VARCHAR2(1)
);
The command from the prompt:
sqlldr parfile=q.par
This test box is driving me nuts. Now I get this error with the sqlldr -
Message 2100 not found; No message file for product=RDBMS, facility=ULMessage 2100 not found; No message file for product=RDBMS
Message 2100 not found; No message file for product=RDBMS, facility=ULMessage 2100 not found; No message file for product=RDBMS
Make sure your ORACLE_HOME is properly set. Typically in the past when I get message not found messages it is from an environment setup issue.
>>How do you handle possible single quotes in the data?
SQL Loader handles them for you.
Look at my example above. The second record I added has a single quote but it's hard to see: Al'fa.
SQL Loader handles them for you.
Look at my example above. The second record I added has a single quote but it's hard to see: Al'fa.
Ah, yeah! So this works but unfortunately when I test it out using a real data file, it only loads 124 of the 91k rows before it errors out with "MAXIMUM ERROR COUNT EXCEEDED - Above statistics reflect partial run." It seems some of the records sent have two tabs where there should be one and data is shifted over for some records. The old Perl script seems to have handled that better.
>>It seems some of the records sent have two tabs where there should be one
Two tabs side by side typically means a null value for a field. This by itself is normal.
>> and data is shifted over for some records
Explain? What it looks like in an editor (like vi) and what it actually is, is two different things.
There should be a 'log' file generated that will tell you why the records failed and a 'bad' file that has the ones that failed.
Two tabs side by side typically means a null value for a field. This by itself is normal.
>> and data is shifted over for some records
Explain? What it looks like in an editor (like vi) and what it actually is, is two different things.
There should be a 'log' file generated that will tell you why the records failed and a 'bad' file that has the ones that failed.
Actually, my mistake. A column was out of place! I fixed that and the data loads in 3 mins 1 second! THANKS!!!
No problem. Glad to help. Told you it was very good at what it does!
You might be able to make it even faster but you need to understand what it actually doing. It isn't for all situations.
Read up on Direct Path loading:
http://docs.oracle.com/cd/E11882_01/server.112/e22490/ldr_concepts.htm#SUTIL997
You might be able to make it even faster but you need to understand what it actually doing. It isn't for all situations.
Read up on Direct Path loading:
http://docs.oracle.com/cd/E11882_01/server.112/e22490/ldr_concepts.htm#SUTIL997
Ok, thank you! :)
I encountered a problem... how do I compare this new table that contains NULLS with another table that has ' ' instead of NULL? When I compare them using MERGE, all 91+k records having their fields with a ' ' value updated with a NULL value. I don't want the field to be updated if it has ' ' and this new field has a NULL.
I encountered a problem... how do I compare this new table that contains NULLS with another table that has ' ' instead of NULL? When I compare them using MERGE, all 91+k records having their fields with a ' ' value updated with a NULL value. I don't want the field to be updated if it has ' ' and this new field has a NULL.
You have lost me.
In Oracle an empty string, '', is technically considered a null.
For example:
select 'hello' from dual where '' is null;
Going from your previous requirements, you wanted a 'None' string inserted when the field was null. My posted test case does this. Look at the NVL function on FIELD2.
You now mention a MERGE statement. Where did this come into the picture?
Can you update your requirements or explain a lot more about what you are wanting to do?
If the loaded data is being loaded into a staging table, you might be better off using External Tables (mentioned above). These use SQL Loader style syntax but allow you to use them like a 'normal' table without having to actually load the data into a 'normal' table.
In Oracle an empty string, '', is technically considered a null.
For example:
select 'hello' from dual where '' is null;
Going from your previous requirements, you wanted a 'None' string inserted when the field was null. My posted test case does this. Look at the NVL function on FIELD2.
You now mention a MERGE statement. Where did this come into the picture?
Can you update your requirements or explain a lot more about what you are wanting to do?
If the loaded data is being loaded into a staging table, you might be better off using External Tables (mentioned above). These use SQL Loader style syntax but allow you to use them like a 'normal' table without having to actually load the data into a 'normal' table.
Maybe this is something new? Should I start a new question with new points?
tab1 - the table populated by the data file with sqlLoader; contains current records
tab2 - existing table of all records - current and historical
I have a script that compares tab1's fields with tab2's field and makes updates to records
in tab2 that have changed.
My tables -
SQL> desc tab1
Name Null? Type
--------------------------
FIELD1 VARCHAR2(9)
FIELD2 VARCHAR2(16)
FIELD3 VARCHAR2(2)
FIELD4 VARCHAR2(64)
FIELD5 VARCHAR2(8)
FIELD6 VARCHAR2(31)
SQL> desc tab2;
Name Null? Type
--------------------------
FIELD1 NOT NULL VARCHAR2(9)
FIELD2 NOT NULL VARCHAR2(16)
FIELD3 NOT NULL VARCHAR2(2)
FIELD4 NOT NULL VARCHAR2(64)
FIELD5 NOT NULL VARCHAR2(8)
FIELD6 NOT NULL VARCHAR2(31)
sql part of my script -
MERGE INTO tab2 a
USING
(select field1,field2,field3,field
minus
select field1,field2,field3,field
ON (a.field2 = b.field2)
WHEN MATCHED THEN
UPDATE SET
a.field1 = b.field1,
a.field3 = b.field3,
a.field4 = b.field4,
a.field5 = b.field5,
a.field6 = b.field6
The error message:
DBD::Oracle::db do failed: ORA-01407: cannot update ("TAB2"."FIELD1") to NULL
tab1 - the table populated by the data file with sqlLoader; contains current records
tab2 - existing table of all records - current and historical
I have a script that compares tab1's fields with tab2's field and makes updates to records
in tab2 that have changed.
My tables -
SQL> desc tab1
Name Null? Type
--------------------------
FIELD1 VARCHAR2(9)
FIELD2 VARCHAR2(16)
FIELD3 VARCHAR2(2)
FIELD4 VARCHAR2(64)
FIELD5 VARCHAR2(8)
FIELD6 VARCHAR2(31)
SQL> desc tab2;
Name Null? Type
--------------------------
FIELD1 NOT NULL VARCHAR2(9)
FIELD2 NOT NULL VARCHAR2(16)
FIELD3 NOT NULL VARCHAR2(2)
FIELD4 NOT NULL VARCHAR2(64)
FIELD5 NOT NULL VARCHAR2(8)
FIELD6 NOT NULL VARCHAR2(31)
sql part of my script -
MERGE INTO tab2 a
USING
(select field1,field2,field3,field
minus
select field1,field2,field3,field
ON (a.field2 = b.field2)
WHEN MATCHED THEN
UPDATE SET
a.field1 = b.field1,
a.field3 = b.field3,
a.field4 = b.field4,
a.field5 = b.field5,
a.field6 = b.field6
The error message:
DBD::Oracle::db do failed: ORA-01407: cannot update ("TAB2"."FIELD1") to NULL
>> Should I start a new question with new points?
Yes. You can post the link to this one in the new one as a reference. The reason is that you asked how to make the load of records faster. That was answered. The merge piece is different.
When you ask the new question please provide sample data (for both tables) and expected results. With what you posted, I believe External Tables is better than SQL Loader into a temp table.
As long as the new question is in the Oracle Zone, I'll see it (as will all the Oracle Experts that monitor that zone).
Yes. You can post the link to this one in the new one as a reference. The reason is that you asked how to make the load of records faster. That was answered. The merge piece is different.
When you ask the new question please provide sample data (for both tables) and expected results. With what you posted, I believe External Tables is better than SQL Loader into a temp table.
As long as the new question is in the Oracle Zone, I'll see it (as will all the Oracle Experts that monitor that zone).
Premium Content
You need an Expert Office subscription to comment.Start Free Trial