Snakemake: strip off path from input - snakemake

Question

I want to write a pipeline in snakemake that takes an input file from config.yaml, runs a command, and writes the output to the current directory under the original filename + new suffix.
Snakefile
configfile: "config.yaml"
rule target:
input:
config["reads"]+".fasta.gz",
rule raw_convert:
input:
config["reads"]
output:
config["reads"]+".fasta.gz" # old path specified here
shell:
"sed -n '1~4s/^#/>/p;2~4p' {input} | gzip > {output}"
config.yaml
reads: /path/to/dir/myreads.fq.gz
Using bash, I would write something like to get the file myreads.fq.gz.fasta.gz:
sed -n '1~4s/^#/>/p;2~4p' ${input} | gzip >$(basename ${input}).fasta.gz

Answer 1

In this solution, I pair read basenames to their full path in a dict, and then use it in rules. This would fail if basenames are not unique though.
import os
d = {}
for read in config["reads"]:
basename = os.path.basename(read)
d[basename] = read
rule all:
input:
expand('{read_basename}.fasta.gz', read_basename=list(d.keys()))
rule xxx:
input:
lambda wildcards: d[wildcards.read_basename]
output:
"{read_basename}.fasta.gz"
shell:
'soemthing'
You may want to replace .fq.gz with .fasta.gz instead of appending them. For readability purposes.

Answer 2

I finally came up with some code that seems to do the trick:
configfile: "config.yaml"
import os
basenamereads = os.path.basename(config["reads"])
rule target:
input: expand("{myoutput}.fasta.gz", myoutput=basenamereads)
rule xxx:
input:
config["reads"]
output:
os.path.basename(config["reads"])+".fasta.gz"
shell:
"cat {input} >{output}"