Reproduce the issue
If you have an issue with usearch
that you want me to look at, please make a toy dataset that
reliably reproduces the issue and a runme.bash script in a standalone
directory that executes the usearch command(s) starting from clean
data files (usually, FASTA). Toy
dataset
The toy dataset should be as small and simple as
possible. This makes the problem easier to analyze and is more
convenient, e.g. should ideally be small enough to send as an email
attachment.
Eliminate
unnecessary commands
The example should reduced to a single usearch command.
Isolate the command which causes the problem by testing the commands one
at a time.
Reduce data size
Reduce the data size as much as possible. This is helpful even if
your dataset is small to start with. There are many ways to do this,
e.g.:
- the
fastx_subsample command can
be used to select a small subset from a large file.
- Binary search: divide the input
data into two and run usearch on each half separately. Repeat until you no
longer see the problem.
- Cluster
the input at some low identity, say 80%, then try running usearch the
sequences in one of the clusters.
Create a standalone
directory
Configure your toy dataset so
that all the input files are in a single directory. The goal of this
step is to enable a runme script that does not use path names so can be
installed anywhere in a file system.
Make a runme script
Make a script named runme.bash
for the bash
shell with the usearch command(s) that reproduce your issue.
Ideally, the runme script should be a single usearch command line.
If this is not possible, then the number of calls to usearch should be
reduced as far as possible and other commands (e.g., standard Linux
commands) should also be reduced as much as possible. There should be no
paths in filenames used in the script.
Package the standalone
directory
Make a standalone directory that
contains: (1) the input files for usearch, typically this will be one or
a few FASTA files, (2) the runme.bash script, and (3) the output files
you get, so that I can verify I get the same results. Create a gzip
compressed tar file using a command like this:
tar -zcvf /tmp/toy.tz
.
Note the period (.) at the end of the
command-line above, and note also that the toy.tz file should be created
in /tmp or some other directory, not in the stand-alone directory.
Verify the toy works
correctly
Create a new directory, extract
toy.tz into that directory, and verify that running ./runme.bash
reproduces your issue.
Send the toy to me
If the script is small enough,
send as an email attachment. Otherwise, you can use a free
file sharing service. Occasionally, email systems filter messages
with attachments that contain executable files such as scripts, even if
they are embedded in tarball files like toy.tz. As a precaution, it is a
good idea to send a second email to confirm that the attachment has been
send.
Other information
Please also send the following
information:
- Usearch version information. This
should include the "three-dot" version number e.g. 5.2.13,
the "bitness" (32 or 64) and the binary format (win32,
win64, linux or osx).
- Platform. Send the output from uname
-a.
|