scilight

SciLight - Simple task execution in Python scripts

A super-simple library for performing stepwise batch tasks (implemented as shell commands, or python functions) that saves things to files, such that outputs from already finished tasks are not needlessly re-computed. See the below for an example.

SciLight does not (currently) have a scheduler or central worker pool or anything like that. Instead you simply execute your tasks manually in a procedural way. This way task executions can easily be mixed with other procedural python code.

SciLight can work as an alternative to full-blown workflow frameworks like Luigi or Airflow for cases when you just have a single python script, where you want to do a few batch steps before starting your interactive analysis, such as downloading datasets, unpacking them, preprocessing et cetera.

SciLight is small (not much more than 100 lines of code), and has no external dependencies, meaning that you can even copy the implementation into your own code repos if you want to ensure maximum future reproducibility.

What does 'SciLight' mean?

SciLight is named as such, as being a lighter version of SciPipe, also written by the author, in Go. If you need true multicore-performance and a compiled language, you might want to have a look at SciPipe instead.

Prerequisites

• SciLight is so far only tested on unix-like environments (it is runnable on Windows via Windows Sybsystem for Linux (WSL)).

Installation

Install from the Python Package Index using pip:

pip install scilight

Usage

SciLight works by specifying either a shell command, or a python function to be executed, as the first argument to scilight.shell() or scilight.func() respectively.

In shell commands, you need to replace input and output file paths with placeholders on the form of [i:inputname] and [o:outputname:outputpath] respectively. Additionally, you can provide parameters (as string values) using the [p:paramname] syntax. You will also need to provide dicts which specify the paths to the inputs, outputs and parameters as appropriate, by providing them to the optional inputs, outputs and params parameters of scilight.shell() and scilight.func(). See the example below for a concrete example.

Inputs should always be provided via the inputs-parameter, while output paths are easiest to provide inline in the command, in the respective placeholder. Note that you can re-use input placeholder values to produce the output path. So, for example, if you want to name your output the same as the input, but with an extra .txt extension, you can specify it like this in the command: somecommand > [o:myoutput:[i:myinput].txt].

Removing file extensions

If you have an existing extension in the input that you want to remove, you can do it by adding |%.actual-extension-here in the input placeholder. So, if you have an input myinput with the path myfile.txt.gz, you can reuse just the myfile.txt part by writing [i:myinput|%.gz], to remove the .gz part. Putting that inside an output placeholder, you could for example do: zcat [i:archivefile] > [o:unpacked:[i:archivefile|%.gz]], in order to name the unpacked file the same as the archive, but without the .gz extension.

Removing parent directories from paths

Often it is the case that the input path contains a long folder path that you don't want to re-use when re-using the input filename. To clean the path from the parent directory structure, you can add the |basename modifier inside any path placeholder. So, if you have an input myinput with the path some/directory/structure/file.txt.gz, you can reuse just the myfile.txt part by writing [i:myinput|basename], to remove the some/directory/structure part. Modifiers can be compbined, so for example, given that you have an archive file in another directory named some/directory/structure, you could do the following to extract the archive, removing the .gz file extension and putting the extracted file in a new directory named other-directory: zcat [i:archivefile] > other-directory/[o:unpacked:[i:archivefile|basename|%.gz]]

Search and replace in paths

Sometimes you want to clean up paths in more creative ways. This could be for example replacing spaces with _, to avoid problems on Unix-like systems. This can be done inside path formatters by using the format [i:myinput|s/SEARCH/REPLACE/], so, replacing spaces with underscores would be [i:myinput|s/ /_/].

See the example below for how to use some of this in practice!

Example

Below is a small example that downloads a gzipped text file (in the so called FASTA format), un-gzips it, and then calculates the number of A:s, T:s, G:s and C:s and calculates the fraction of G and C:s in relation to all A, T, G, C:s (the so-called GC-fraction measure for DNA).

The two first tasks are done by executing shell commands, and the second one using a python function.

import scilight as sl

# ------------------------------------------------------------------------
# Download a gzipped fasta file and save it as chrmt.fa.gz
# ------------------------------------------------------------------------
url = 'ftp://ftp.ensembl.org/pub/release-100/fasta/'+ \
      'homo_sapiens/dna/Homo_sapiens.GRCh38.dna.chromosome.MT.fa.gz'
download_task = sl.shell(f'wget -O [o:gz:chrmt.fa.gz] {url}')

# ------------------------------------------------------------------------
# Un-GZip the file, into a file named chrmt.fa
# ------------------------------------------------------------------------
ungzip_task = sl.shell('zcat [i:gz] > [o:fa:[i:gz|%.gz]]',
        inputs={'gz': download_task.outputs['gz']})

# ------------------------------------------------------------------------
# Count the fraction of G+C, vs G+C+A+T
# ------------------------------------------------------------------------
# A function for Count GC fraction in DNA
def count_gcfrac_func(task):
    gc_count = 0
    at_count = 0

    with open(task.inputs['fa']) as infile:
        for line in infile:
            if line[0] == '>':
                continue
            for char in line:
                if char in ['A', 'T']:
                    at_count += 1
                elif char in ['G', 'C']:
                    gc_count += 1

    gc_fraction = gc_count/(gc_count+at_count)

    with open(task.outputs['gcfrac'], 'w') as outfile:
        outfile.write(f'{gc_fraction}\n')

# Execute the function
count_task = sl.func(count_gcfrac_func,
        inputs={'fa': ungzip_task.outputs['fa']},
        outputs={'gcfrac': 'gcfrac.txt'})

Add this code to a file named gcfrac.py and run it with:

python gcfrac.py