X-Git-Url: https://git.donarmstrong.com/?p=using_make_for_science.git;a=blobdiff_plain;f=using_make_for_science.Rnw;h=8ff99d10871fb6470e71e3d43a677c3f6518bc0e;hp=79034541d234b605c9e182ba2f66137b58d8dce3;hb=965452df6fbe769c2733543b8880db096a1fb850;hpb=bdd63bfdd95e9ca6308584f316552d891fdd75f8

diff --git a/using_make_for_science.Rnw b/using_make_for_science.Rnw
index 7903454..8ff99d1 100644
--- a/using_make_for_science.Rnw
+++ b/using_make_for_science.Rnw
@@ -17,6 +17,7 @@
 \usepackage{tikz}
 \usepackage{nameref}
 \usepackage{zref-xr,zref-user}
+\usepackage{metalogo}
 \IfFileExists{upquote.sty}{\usepackage{upquote}}{}
 \mode<article>{
   \usepackage[x11names,svgnames,usenames,dvipsnames]{xcolor}
@@ -42,6 +43,10 @@
 }
 
 \usepackage{minted}
+\usepackage{tcolorbox}
+\usepackage{etoolbox}
+\BeforeBeginEnvironment{minted}{\begin{tcolorbox}}%
+\AfterEndEnvironment{minted}{\end{tcolorbox}}%
 
 \mode<presentation>{ 
   \usetheme{CambridgeUS}
@@ -128,12 +133,31 @@
   \item Gradle (Rake DSL)
   \item Leiningen (Clojure)
   \item Tweaker (task definitions in any language)
+  \item Ruffus (Pipeline library for python)
   \item
     \href{https://en.wikipedia.org/wiki/List_of_build_automation_software}{Wikipedia
       List of build automation software}
   \end{itemize}
 \end{frame}
 
+\subsection{Why use GNU make?}
+\begin{frame}{Why use GNU make?}
+  \begin{itemize}
+  \item Ubiquitous -- any machine which you can run command line tools
+    on has GNU make available.
+  \item Large community -- lots of people use GNU make. It's well
+    understood and you can get questions answered
+  \item Simple rules -- all of the rules are in a simple text file
+    which is easily edited and version controlled
+  \item Reasonable debugging -- you can see the commands that make is
+    going to run fairly easily: \mintinline{shell}{make -n tgt;}
+  \item Parallel -- make can make targets in parallel:
+    \mintinline{shell}{make -j8 all;}
+  \item Language agnostic -- make doesn't care what language your code
+    is written in
+  \end{itemize}
+\end{frame}
+
 \section{Introduction to Makefiles}
 
 \begin{frame}[fragile]{Simple Makefile}
@@ -141,46 +165,293 @@
 hello_world:
 	echo "hello world" > hello_world
 \end{minted}
+
 \end{frame}
 
 \subsection{General Syntax}
 
-
 \begin{frame}[fragile]{Simple Makefile}
 \begin{minted}[showtabs]{make}
-hello_world:
-	echo "hello world" > hello_world
+TARGETS: PREREQUISITES
+	RECIPE
 \end{minted}
+\begin{itemize}
+\item TARGETS are file names separated by spaces
+\item PREREQUISITES are file names separated by spaces.
+\item RECIPE lines start with a tab, are executed by the shell and
+  describe how to make the TARGETS (generally from the PREREQUISITES)
+\item A TARGET is out of date if it does not exist or if it is older
+  than any of the prerequisites.
+\end{itemize}
 \end{frame}
 
 \subsection{Variables}
 
+\begin{frame}[fragile]{Some Variables}
+\begin{itemize}
+\item Two flavors of variables
+  \begin{itemize}
+  \item \mintinline{make}{FOO=bar} -- recursively expanded variables;
+    references to other variables are expanded at the time this
+    variable is expanded
+  \item \mintinline{make}{FOO:=bar} -- simply expanded variables; the
+    value is assigned at the moment the variable is created
+  \end{itemize}
+\item Variables can come from the environment and can be overridden on
+  the command line: \mintinline{shell}{make FOO=bleargh} or
+  \mintinline{shell}{FOO=blah make}.
+\item \mintinline{make}{$@} -- tgt name %$
+\item \mintinline{make}{$*} -- current stem %$ 
+\item \mintinline{make}{$^} -- all prerequisites %$
+\item \mintinline{make}{$<} -- first prerequisite %$
+\item \mintinline{make}{$(FOO)} -- how variables are referenced %$
+\end{itemize}
+\end{frame}
+
+\subsection{Functions}
+
+\begin{frame}[fragile]{Some Functions}
+  \begin{itemize}
+  \item \mintinline{make}{$(patsubst %.bam,%.sam,foo.sam bar.sam)} %$
+      -- returns foo.bam bar.bam.
+  \item \mintinline{make}{$(filter-out %.bam,foo.sam bar.bam)} %$
+      -- returns foo.sam
+  \item \mintinline{make}{$(words foo.sam bar.bam)} %$
+      -- returns the number of words in its argument (2)
+  \item \mintinline{make}{$(wordlist 1,2,foo.sam bar.bam bleargh.foo)} %$
+      -- returns the words in its last argument starting with the 1st
+      and ending with the second.
+  \end{itemize}
+\end{frame}
+
 \subsection{Rules}
 
 \subsubsection{Default Target}
 
+\begin{frame}[fragile]{How does make know what to build?}
+\begin{minted}[showtabs]{make}
+first_tgt: 
+	touch $@
+second_tgt: first_tgt
+	touch $@
+\end{minted}
+  \begin{itemize}
+  \item By default, make builds the first tgt.
+  \item You can specify a specific tgt to build on the command line
+    (\mintinline{shell}{make first_tgt}).
+  \item You can change the default tgt by using the variable
+    \mintinline{make}{.DEFAULT_GOAL := second_tgt}
+  \end{itemize}
+\end{frame}
+
+
 \subsubsection{Special Targets}
 
+\begin{frame}[fragile]{Special Targets}
+\begin{minted}[showtabs]{make}
+.PHONY: clean
+
+clean:
+	rm -f first_tgt second_tgt
+\end{minted}
+  \begin{itemize}
+  \item \mintinline{make}{.PHONY} -- any time make considers this
+    tgt, it is run unconditionally, even if a file exists.
+  \item \mintinline{make}{.ONESHELL} -- when a tgt is built, all
+    lines will be given to a single invocation of the shell.
+  \item Lots of other special targets which are not described here.
+  \end{itemize}
+\end{frame}
+
+
 \subsubsection{Pattern Rules}
 
-\subsection{Functions}
 
+\begin{frame}[fragile]{Special Targets}
+\begin{minted}[showtabs]{make}
+%.fasta.gz: %.fasta
+	gzip -dc $< > $@
+
+%.bam: %.sam
+	samtools view -b -o $@ $<
+\end{minted}
+  \begin{itemize}
+  \item \% is the pattern stem which is accessible by
+    \mintinline{make}{$*} %$ within rules
+  \item The first rule uncompresses fasta files
+  \item The second rule turns sam files into bam files
+  \end{itemize}
+\end{frame}
 
 \section{Examples}
 
 \subsection{This Presentation}
 
-\subsection{Can you dig it?}
+\begin{frame}[fragile]{How this presentation is made}
+\inputminted[showtabs]{make}{Makefile}
+  \begin{itemize}
+  \item all is the default tgt
+  \item Make .tex files from the knitr source.
+  \item The third rule uses latexmk to build the pdf using \XeLaTeX.
+  \end{itemize}
+\end{frame}
 
 \subsection{Calling records from SRA}
 
+\begin{frame}{Calling records from SRA: The problem}
+  \begin{itemize}
+  \item ≈200 tissue samples from Roadmap Epigenomics
+  \item No consistent workflow
+  \item Reanalyze them all using STAR and cufflinks
+  \end{itemize}
+\end{frame}
+
+\begin{frame}[fragile]{Calling records from SRA: Downloading}
+\begin{minted}[showtabs,breaklines]{make}
+SRX=SRX007165
+SRRS=SRR020291 SRR020290
+NREADS=1
+SRR_FILES=$(patsubst %,%.sra,$(SRRS))
+
+get_srr: $(SRR_FILES)
+
+$(SRR_FILES): %.sra:
+	rsync -avP "rsync://ftp-trace.ncbi.nlm.nih.gov/sra/sra-instant/reads/ByRun/sra/SRR/$(shell echo -n $*|sed 's/\(SRR[0-9][0-9][0-9]\).*/\1/')/$*/$*.sra" $@;
+\end{minted}
+%$
+  \begin{itemize}
+  \item First three lines are actually generated by other code and
+    included
+  \item Download all of the SRA files using rsync
+  \end{itemize}
+\end{frame}
+
+\begin{frame}[fragile]{Calling records from SRA: Dumping fastq}
+\begin{minted}[showtabs,breaklines]{make}
+ifeq ($(NREADS),1)
+FASTQ_FILES:=$(patsubst %,%.fastq.gz,$(SRRS))
+else
+FASTQ_FILES:=$(patsubst %,%_1.fastq.gz,$(SRRS))  $(patsubst %,%_2.fastq.gz,$(SRRS))
+endif
+
+make_fastq: $(FASTQ_FILES)
+
+ifeq ($(NREADS),1)
+$(FASTQ_FILES): %.fastq.gz: %.sra
+else
+%_1.fastq.gz %_2.fastq.gz: %.sra
+endif
+	$(MODULE) load sratoolkit/2.3.5-2; \
+	fastq-dump --split-3 --gzip $^;
+\end{minted}
+%$
+  \begin{itemize}
+  \item Call fastq-dump to dump the fastq files
+  \item Handles NREADS of 1 and 2 differently
+  \end{itemize}
+\end{frame}
+
+\begin{frame}[fragile]{Calling records from SRA: Align with STAR}
+\begin{minted}[showtabs,breaklines]{make}
+$(SRX)_star.bam:
+	$(MODULE) load STAR/2.4.2a; \
+	mkdir -p $(SRX)_star; \
+	STAR --outFileNamePrefix $(SRX)_star/ \
+	--outSAMtype BAM SortedByCoordinate \
+	--runThreadN $(CORES) \
+	--outSAMstrandField intronMotif \
+	--genomeDir $(STAR_INDEX_DIR) \
+	--readFilesCommand "gzip -dc" \
+	--readFilesIn $(TOPHAT_FASTQ_ARGUMENT);
+	ln $(SRX)_star/Aligned.*.bam $@ -s
+\end{minted}
+%$
+  \begin{itemize}
+  \item Call STAR with lots of options to do the alignment
+  \end{itemize}
+\end{frame}
+
+\begin{frame}[fragile]{Calling records from SRA: Call with cufflinks}
+\begin{minted}[showtabs,breaklines]{make}
+call: $(SRX)_genes.fpkm_tracking
+
+$(SRX)_genes.fpkm_tracking: $(SRX)_star.bam $(BOWTIE_INDEX_DIR)$(GTF)
+	$(MODULE) load cufflinks/2.2.1; \
+	cufflinks -p $(CORES) -G $(wordlist 2,2,$^) $<
+	for file in genes.fpkm_tracking isoforms.fpkm_tracking skipped.gtf transcripts.gtf; do \
+		mv $${file} $(SRX)_$${file}; \
+	done;
+\end{minted}
+%$
+  \begin{itemize}
+  \item Use cufflinks to call
+  \end{itemize}
+\end{frame}
+
+\begin{frame}[fragile]{Run it on biocluster}
+  \begin{minted}{shell}
+~donarm/uiuc_igb_scripts/dqsub --mem 70G \
+   --ppn 8 make call;
+\end{minted}
+\begin{itemize}
+\item dqsub is my own qsub wrapper which avoids me having to write
+  little scripts for everything
+\item \url{http://git.donarmstrong.com/?p=uiuc_igb_scripts.git;a=blob;f=dqsub}
+\end{itemize}
+\end{frame}
+
 \section{Why not make?}
 
+\begin{frame}{Why not make?}
+  \begin{itemize}
+  \item Timestamps, not MD5sums
+  \item Complicated workflows
+  \end{itemize}
+\end{frame}
+
 \subsection{Timestamps}
 
+\begin{frame}[fragile]{Dealing with timestamps}
+  \begin{minted}[showtabs,breaklines]{make}
+TARGET: PREREQ1 PREREQ1
+	if [ -e $@.tgt.md5sum ] && [ -e $@ ] \
+	   && md5sum --status --check \
+	   $@.tgt.md5sum; then \
+	 touch $@; \
+	else \
+	 RECIPE FOR $@; \
+	 md5sum $^ > $@.tgt.md5sum; \
+	fi;
+\end{minted}
+% $
+\begin{itemize}
+\item Make builds things on the basis of timestamps
+\item But what if the contents haven't changed and it's expensive to
+  rebuild?
+\item Use md5sum!
+\end{itemize}
+\end{frame}
+
 \subsection{Complicated Workflows}
 
+\begin{frame}{What about complicated workflows?}
+  \begin{itemize}
+  \item If your workflow is really complicated, what then?
+  \item Use some other language to write your workflow in
+  \item Use a simple makefile which just runs the workflow
+  \end{itemize}
+\end{frame}
+
 \section{Further Resources}
 
+\begin{frame}{Further Resources}
+  \begin{itemize}
+  \item GNU Make Manual: \url{https://www.gnu.org/software/make/manual/}
+  \item Mailing lists: \url{http://www.gnu.org/software/make/}
+  \item Stack overflow: \url{http://stackoverflow.com/questions/tagged/make}
+  \item Myself: \href{mailto:don@donarmstrong.com}{don@donarmstrong.com}
+  \item This presentation: \url{http://git.donarmstrong.com/using\_make\_for\_science.git}
+  \end{itemize}
+\end{frame}
 
 \end{document}