Dear All,
I would like to split a file of the following format into multiple files based on the number in the 6th column (numbers 1, 2, 3...):
ATOM 1 N GLY A 1 -3.198 27.537 -5.958 1.00 0.00 N
ATOM 2 CA GLY A 1 -2.199 28.399 -6.617 1.00 0.00 ... (3 Replies)
Hello,
What's the best way to split a large into multiple files based on the last digit in the first column.
input file:
f
2738483300000x0y03772748378831x1y13478378358383x2y23743878383802x3y33787828282820x4y43748838383881x5y5
Desired Output:
f0
3738483300000x0y03787828282820x4y4
f1... (9 Replies)
I am unable to spit the file based on the 2nd column passing as a parameter with awk command.
Source file:
“100”,”customer information”,”10000”
“200”,”customer information”,”50000”
“300”,”product information”,”40000”
script: the command is not allowing to pass the parameters with the awk... (7 Replies)
Hi,
I have a fixed width text file without any header row. One of the columns contains a date in YYYYMMDD format.
If the original file contains 3 dates, I want my shell script to split the file into 3 small files with data for each date.
I am a newbie and need help doing this. (14 Replies)
Hi All
I have one query,say i have a requirement like the below code should be
move to diffent files whose maximum lines can be of 10 lines.Say in the below example,it consist of 14 lines.
This should be moved logically using the data in the fisrt coloumn to file1 and file 2.The data of first... (2 Replies)
i have file1.txt
asdas|csada|130310|0423|A1|canberra
sdasd|sfdsf|130426|2328|A1|sydney
Expected output : on eaceh third and fourth colum, split into each two characters
asdas|csada|13|03|10|04|23|A1|canberra
sdasd|sfdsf|13|04|26|23|28|A1|sydney (10 Replies)
Hi,
I have a similar input format-
A_1 2
B_0 4
A_1 1
B_2 5
A_4 1
and looking to print in this output format with headers. can you suggest in awk?awk because i am doing some pattern matching from parent file to print column 1 of my input using awk already.Thanks!
letter number_of_letters... (5 Replies)
Hi All,
I have a requirement to split file into 2 sets of file. Below is a sample data of the file
AU;PTN;24EX;25-AUG-14;AU;123;SE;123;Test NN;;;;ASD;
AU;PTN;24EX;25-AUG-14;AU;456;SE;456;Test NN;;;;ASD;
AU;PTN;24EX;25-AUG-14;AU;147;SE;147;Test NN;;;;ASD;... (6 Replies)
Hi all,
Newbie here, so please bear over with my stupid question :)
I have used far too long time today on figuring this out, so I hope that someone here can help me move on.
I have some annotation data for a transcriptome where I want to split a column containing NCBI accession IDs into a... (7 Replies)
Hi Team,
I have a requirement in such a way that need to split the file into two based on which column particular value appears.Please find my sample file below.
Lets consider the delimiter of this file as either comma or two colons.(:: and ,). So I need to split the file in such a way that all... (2 Replies)
Discussion started by: ginrkf
2 Replies
LEARN ABOUT DEBIAN
io::handle::prototype::fallback
IO::Handle::Prototype::Fallback(3pm) User Contributed Perl Documentation IO::Handle::Prototype::Fallback(3pm)NAME
IO::Handle::Prototype::Fallback - Create IO::Handle like objects using a set of callbacks.
SYNOPSIS
my $fh = IO::Handle::Prototype::Fallback->new(
getline => sub {
my $fh = shift;
...
},
);
DESCRIPTION
This class provides a way to define a filehandle based on callbacks.
Fallback implementations are provided to the extent possible based on the provided callbacks, for both writing and reading.
SPECIAL CALLBACKS
This class provides two additional methods on top of IO::Handle, designed to let you implement things with a minimal amount of baggage.
The fallback methods are all best implemented using these, though these can be implemented in terms of Perl's standard methods too.
However, to provide the most consistent semantics, it's better to do this:
IO::Handle::Prototype::Fallback->new(
__read => sub {
shift @array;
},
);
Than this:
IO::Handle::Prototype::Fallback->new(
getline => sub {
shift @array;
},
);
Because the fallback implementation of "getline" implements all of the extra crap you'd need to handle to have a fully featured
implementation.
__read
Return a chunk of data of any size (could use $/ or not, it depends on you, unlike "getline" which probably should respect the value of
$/).
This avoids the annoying "substr" stuff you need to do with "read".
__write $string
Write out a string.
This is like a simplified "print", which can disregard $, and "$" as well as multiple argument forms, and does not have the extra
"substr" annoyance of "write" or "syswrite".
WRAPPING
If you provide a single reading related callback ("__read", "getline" or "read") then your callback will be used to implement all of the
other reading primitives using a string buffer.
These implementations handle $/ in all forms ("undef", ref to number and string), all the funny calling conventions for "read", etc.
FALLBACKS
Any callback that can be defined purely in terms of other callbacks in a way will be added. For instance "getc" can be implemented in terms
of "read", "say" can be implemented in terms of "print", "print" can be implemented in terms of "write", "write" can be implemented in
terms of "print", etc.
None of these require special wrapping and will always be added if their dependencies are present.
GLOB OVERLOADING
When overloaded as a glob a tied handle will be returned. This allows you to use the handle in Perl's IO builtins. For instance:
my $line = <$fh>
will not call the "getline" method natively, but the tied interface arranges for that to happen.
perl v5.10.1 2009-09-29 IO::Handle::Prototype::Fallback(3pm)