it will solve lookup and replicate just write query yes/no

to find records which should be inserted , updated or
deleted one should use ab initio flow 
a. unload master table
b. read delta file 
c. use inner join to join a and b unused a will be your
delete records (if required) unused b will be your insert
record . joined a and b will be your update records

to view MFS in unix you should run m_expand command
conditional DML can be pass as program variable 
conditional components will be used only when condition past
to the graph is true

In banking and financial modules every file will consists 
of Only one header records & Multiple Trailers (Transaction 
trailer & Batch trailer.

If it contains junk data we can do it by using Sed Editor 
or With help of PERAL SCRIPT.

1. If you know the signature of header and tailer record 
then use filerby expression component to filter the header 
and tailer records

2. Use one reformate component and then inside the 
transformation use next_in_sequence() function to assign 
unique numbers to each record,and then use filter by 
expression component to filter the records based on 
sequence numbers.

3.Follow the step 2 and use instead of filter by expression 
component use leading records component to filter the 
header and tailer records.

Take 10,000 records as source file and take output table as 
another source and then Join both the source,select input 
must be sorted parameter option in the Join component,If 
any matching records found then put that into a thrash, 
take all unmatched records and inert into the target table.

There are several resons that graph will be failed.
I have one specific Answar for this is...

If the graph is failed then Abinitio will create one .rec 
file in the run directory of your sendbox. if you want to 
rollback the graph then use m_rollback command in the unix 
directory or you can use m_cleanup utilities in the Unix 
command.

please let me know your exact requirement so that I can try 
to help you.

When set to In memory: Input need not be sorted, the 
component accepts unsorted input and permits the use of the 
maintain-order parameter. Component sorts records 
internally in its buffer.

When set to Inputs must be sorted, the component requires 
sorted input, and the maintain-order parameter is not 
available.

Default is Inputs must be sorted.

The Inmemory sort and input must be sorted options are 
there in the Join,Rollup and Dedup components.
Main difference between these two is if you selected input 
must be sorted options in the above mentioned components 
the the downstream components will get the records in a 
sorted oder. if you are selected option as Inmemory sort 
then the downstream components will not get the sorted 
records.

1. Start script
2. Graph components.
3.End script

.air-project-parameters
Contains the parameter definitions of all the parameters 
within a sandbox. This file is maintained by the GDE and 
the Ab Initio environment scripts. 

.air-sandbox-overrides 
This file exists only if you are using version 1.11 or a 
later version of the GDE. It contains the user's private 
values for any parameters in .air-project-parameters that 
have the Private Value flag set. It has the same format as 
the .air-project-parameters file.

When you edit a value (in GDE) for a parameter that has the 
Private Value flag checked, the value is stored in the .air-
sandbox-overrides file rather than the .air-project-
parameters file.

In case of inner-join all the records not matching the key 
specified goes to the respective unused ports, in full 
outer-join none of the records goes to the unused ports. In 
case of reject port, records which do not match with DML 
come to the reject port.

In case of inner-join all the records not matching the key 
specified goes to the respective unused ports, in full 
outer-join none of the records goes to the unused ports.
All the records which evaluates to NULL during joiin
transformation will go into reject port if the limit +
ramp*number_of_input_records_so_far <
number_of_input_records_so_far.

MAX_CORE is a temporary memory of component and it is used 
to sort the records and it read and write the files to disk.
-> The perameter avilable when the "SORTED_INPUT" perameter 
is set to IN_MEMORY:"Input Need Not be sorted"
->It contains default values is 67108864(64mb)

NOTE: This perameter avile in JOIN,SORT and 
Assign_Key...etc Components.

Thx
HariBabu

The max-core parameter (for those components that have it) 
allows you to specify the maximum amount of memory that can 
be allocated for the component. 
If the component has enough memory allocated to it, it can 
do everything it needs to do in memory, without having to 
write anything temporarily to disk.

Reads an input record from the in port. 
Passes the input record to the get_filename transform 
function. 
If the get_filename function returns a string that does not 
contain a valid filename, or if the file derived from the 
string is empty or missing, Read Multiple Files proceeds 
according to what is specified in the filename-error, file-
empty, and file-missing parameters. 
If the get_filename function returns a string containing a 
valid filename, Read Multiple Files retrieves the specified 
file. 
 NOTE: The input file specified in the string must be local 
to the machine where Read Multiple Files is running. Note 
also that in multifile layouts, different instances of Read 
Multiple Files can be running on different computers, so 
input filenames referring to directories that exist only on 
particular computers must be partitioned carefully. To use 
relative paths, use the DML function this_partition_path. 

Opens the specified file and reads records from it: 
If an input_type data type is defined in the transform 
package, Read Multiple Files uses this type to read records 
from the file. If you omit input_type, Read Multiple Files 
uses the record type defined on the out port. 
If the filter parameter was set to Range, and the skip-
count parameter is non-0, Read Multiple Files skips the 
specified number of records before reading subsequent 
records from the file. 
If the filter parameter was set to Range, and the read-
count parameter is non-0, Read Multiple Files reads up to 
this maximum number of records from the file. If read_count 
is less than or equal to zero, Read Multiple Files reads 
the remainder of the file. 
Read Multiple Files writes records to the out port. 
If a reformat function is defined in the transform package, 
the component passes each record read from the input file 
to the reformat function, along with the filename. The 
output of the reformat function is written to the out port. 
If you omit a reformat function, but provide an input_type 
data type in the package, the component behaves like a 
Reformat component with no transform package, reading 
records from the file using the input_type, and performing 
any conversion necessary to match the record format on the 
out port. 
If the reject and error ports are connected to flows and 
the reformat transform function returns NULL, Read Multiple 
Files writes an error message to the error port and the 
current record to the reject port. 
Read Multiple Files stops execution of the graph when the 
number of reject events exceeds the value given by the 
following formula:

limit + (ramp * number_of_records_processed_so_far) 

For details, see "Setting limit and ramp for reject events"

If you do not connect flows to either the reject or error 
port, Read Multiple Files does not write an error message 
and discards the rejected record.

There can be many ways to remove header & trailer record 
depending on the data you are receiving. If you have a 
particular signature for header & Trailer record you can 
use Filter by Expression or Partition By Expression 
Component to remove the header & Trailer record. If you 
don't have signature for your header & Trailer record then 
you have to use reformat component to select only the data 
record.

if there is no clue about how the header and trailer would 
be then put a filter with next_in_seq = 1 to remove the 
header and dedup the whole input with all fields as key and 
keep last to remove the trailer.

Yes can be done with the help of REFORMATE component in 
this use reformat out put index.if we use this we can get 
what ever out we need.

OR
With the help of Run progame compnent.

Please paste the error you are getting so that we can 
provide the solution for that.

Regards,
Dhirendra

If these figures are being added, perhaps only 2 bytes was 
allowed for the result instead of 3.

Yes its possible. The Multifile system depth of your system 
depends on the Number of CPU's you have. If you have 1 CPU 
then the max depth of your MFS partition can be 4. In this 
case also you can create 2 way partitions.

Regards,
Dhirenda