Introduction ============ This description is for *creators* of the reproduction pipeline. See `README` for instructions on running it (in short, just download/clone it, then run `./configure` and `./.local/bin/make -j8`). This project contains a **fully working template** for a high-level research reproduction pipeline, or reproducible paper, as defined in the link below. If the link below is not accessible at the time of reading, please see the appendix at the end of this file for a portion of its introduction. Some [slides](http://akhlaghi.org/pdf/reproducible-paper.pdf) are also available to help demonstrate the concept implemented here. http://akhlaghi.org/reproducible-science.html This template is created with the aim of supporting reproducible research by making it easy to start a project in this framework. As shown below, it is very easy to customize this template reproducible paper pipeline for any particular research/job and expand it as it starts and evolves. It can be run with no modification (as described in `README`) as a demonstration and customized for use in any project as fully described below. The pipeline will download and build all the necessary libraries and programs for working in a closed environment (highly independent of the host operating system) with fixed versions of the necessary dependencies. The tarballs for building the local environment are also collected in a [separate repository](https://gitlab.com/makhlaghi/reproducible-paper-dependencies). The [final reproducible paper output](https://gitlab.com/makhlaghi/reproducible-paper-output/raw/master/paper.pdf) of this pipeline is also present in [a separate repository](https://gitlab.com/makhlaghi/reproducible-paper-output). Notice the last paragraph of the Acknowledgements where all the dependencies are mentioned with their versions. Below, we start with a discussion of why Make was chosen as the high-level language/framework for this research reproduction pipeline and how to learn and master Make easily (and freely). The general architecture and design of the pipeline is then discussed to help you navigate the files and their contents. This is followed by a checklist for the easy/fast customization of this pipeline to your exciting research. We continue with some tips and guidelines on how to manage or extend the pipeline as your research grows based on our experiences with it so far. The main body concludes with a description of possible future improvements that are planned for the pipeline (but not yet implemented). As discussed above, we end with a short introduction on the necessity of reproducible science in the appendix. Please don't forget to share your thoughts, suggestions and criticisms on this pipeline. Maintaining and designing this pipeline is itself a separate project, so please join us if you are interested. Once it is mature enough, we will describe it in a paper (written by all contributors) for a formal introduction to the community. Why Make? --------- When batch processing is necessary (no manual intervention, as in a reproduction pipeline), shell scripts are usually the first solution that come to mind. However, the inherent complexity and non-linearity of progress in a scientific project (where experimentation is key) make it hard to manage the script(s) as the project evolves. For example, a script will start from the top/start every time it is run. So if you have already completed 90% of a research project and want to run the remaining 10% that you have newly added, you have to run the whole script from the start again. Only then will you see the effects of the last new steps (to find possible errors, or better solutions and etc). It is possible to manually ignore/comment parts of a script to only do a special part. However, such checks/comments will only add to the complexity of the script and will discourage you to play-with/change an already completed part of the project when an idea suddenly comes up. It is also prone to very serious bugs in the end (when trying to reproduce from scratch). Such bugs are very hard to notice during the work and frustrating to find in the end. The Make paradigm, on the other hand, starts from the end: the final *target*. It builds a dependency tree internally, and finds where it should start each time the pipeline is run. Therefore, in the scenario above, a researcher that has just added the final 10% of steps of her research to her Makefile, will only have to run those extra steps. With Make, it is also trivial to change the processing of any intermediate (already written) *rule* (or step) in the middle of an already written analysis: the next time Make is run, only rules that are affected by the changes/additions will be re-run, not the whole analysis/pipeline. This greatly speeds up the processing (enabling creative changes), while keeping all the dependencies clearly documented (as part of the Make language), and most importantly, enabling full reproducibility from scratch with no changes in the pipeline code that was working during the research. This will allow robust results and let the scientists get to what they do best: experiment and be critical to the methods/analysis without having to waste energy and time on technical problems that come up as a result of that experimentation in scripts. Since the dependencies are clearly demarcated in Make, it can identify independent steps and run them in parallel. This further speeds up the processing. Make was designed for this purpose. It is how huge projects like all Unix-like operating systems (including GNU/Linux or Mac OS operating systems) and their core components are built. Therefore, Make is a highly mature paradigm/system with robust and highly efficient implementations in various operating systems perfectly suited for a complex non-linear research project. Make is a small language with the aim of defining *rules* containing *targets*, *prerequisites* and *recipes*. It comes with some nice features like functions or automatic-variables to greatly facilitate the management of text (filenames for example) or any of those constructs. For a more detailed (yet still general) introduction see the article on Wikipedia: https://en.wikipedia.org/wiki/Make_(software) Many implementations of Make exist and all should be usable with this pipeline. This pipeline has been created and tested mainly with GNU Make which is the most common implementation. But if you see parts specific to GNU Make, please inform us to correct it. How can I learn Make? --------------------- The best place to learn Make from scratch is the GNU Make manual. It is an excellent and non-technical (in its first chapters) book to help get started. It is freely available and always up to date with the current release. It also clearly explains which features are specific to GNU Make and which are general in all implementations. So the first few chapters regarding the generalities are useful for all implementations. The first link below points to the GNU Make manual in various formats and in the second, you can download it in PDF (which may be easier for a first time reading). https://www.gnu.org/software/make/manual/ https://www.gnu.org/software/make/manual/make.pdf If you use GNU Make, you also have the whole GNU Make manual on the command-line with the following command (you can come out of the "Info" environment by pressing `q`). ```shell $ info make ``` If you aren't familiar with the Info documentation format, we strongly recommend running `$ info info` and reading along. In less than an hour, you will become highly proficient in it (it is very simple and has a great manual for itself). Info greatly simplifies your access (without taking your hands off the keyboard!) to many manuals that are installed on your system, allowing you to be much more efficient as you work. If you use the GNU Emacs text editor (or any of its variants), you also have access to all Info manuals while you are writing your projects (again, without taking your hands off the keyboard!). Published works using this pipeline ----------------------------------- The links below will guide you to some of the works that have already been published using the method of this pipeline. Note that this pipeline is evolving, so some small details may be different in them, but they can be used as a good working model to build your own. - Section 7.3 of Bacon et al. ([2017](http://adsabs.harvard.edu/abs/2017A%26A...608A...1B), A&A 608, A1): The version controlled reproduction pipeline is available [on Gitlab](https://gitlab.com/makhlaghi/muse-udf-origin-only-hst-magnitudes) and a snapshot of the pipeline along with all the necessary input datasets and outputs is available in [zenodo.1164774](https://doi.org/10.5281/zenodo.1164774). - Section 4 of Bacon et al. ([2017](http://adsabs.harvard.edu/abs/2017A%26A...608A...1B), A&A, 608, A1): The version controlled reproduction pipeline is available [on Gitlab](https://gitlab.com/makhlaghi/muse-udf-photometry-astrometry) and a snapshot of the pipeline along with all the necessary input datasets is available in [zenodo.1163746](https://doi.org/10.5281/zenodo.1163746). - Akhlaghi & Ichikawa ([2015](http://adsabs.harvard.edu/abs/2015ApJS..220....1A), ApJS, 220, 1): The version controlled reproduction pipeline is available [on Gitlab](https://gitlab.com/makhlaghi/NoiseChisel-paper). This is the very first (and much less mature) implementation of this pipeline: the history of this template pipeline started more than two years after that paper was published. It is a very rudimentary/initial implementation, thus it is only included here for historical reasons. However, the pipeline is complete and accurate and uploaded to arXiv along with the paper. See the more recent implementations if you want to get ideas for your version of this pipeline. Citation -------- A paper will be published to fully describe this reproduction pipeline. Until then, if this pipeline is useful in your work, please cite the paper that implemented the first version of this pipeline: Akhlaghi & Ichikawa ([2015](http://adsabs.harvard.edu/abs/2015ApJS..220....1A), ApJS, 220, 1). The experience gained with this template after several more implementations will be used to make this pipeline robust enough for a complete and useful paper to introduce to the community afterwards. Also, when your paper is published, don't forget to add a notice in your own paper (in coordination with the publishing editor) that the paper is fully reproducible and possibly add a sentence or paragraph in the end of the paper shortly describing the concept. This will help spread the word and encourage other scientists to also publish their reproduction pipelines. Reproduction pipeline architecture ================================== In order to adopt this pipeline to your research, it is important to first understand its architecture so you can navigate your way in the directories and understand how to implement your research project within its framework. But before reading this theoretical discussion, please run the pipeline (described in `README`: first run `./configure`, then `./.local/bin/make -j8`) without any change, just to see how it works. In order to obtain a reproducible result it is important to have an identical environment (for example same versions the programs that it will use). This also has the added advantage that in your separate research projects, you can use different versions of a single software and they won't interfere. Therefore, the pipeline builds its own dependencies during the `./configure` step. Building of the dependencies is managed by `reproduce/src/make/dependencies-basic.mk` and `reproduce/src/make/dependencies.mk`. These Makefiles are called by the `./configure` script. The first is intended for downloading and building the most basic tools like GNU Bash, GNU Make, and GNU Tar. Therefore it must only contain very basic and portable Make and shell features. The second is called after the first, thus enabling usage of the modern and advanced features of GNU Bash and GNU Make, similar to the rest of the pipeline. Later, if you add a new program/library for your research, you will need to include a rule on how to download and build it (in `reproduce/src/make/dependencies.mk`). After configuring, the `.local/bin/make` command will start the processing with the custom version of Make that was locally installed during configuration. The first file that is read is the top-level `Makefile`. Therefore, we'll start our navigation/discussion with this file. This file is relatively short and heavily commented so hopefully the descriptions in each comment will be enough to understand the general details. As you read this section, please also look at the contents of the mentioned files and directories to fully understand what is going on. Before starting to look into the top `Makefile`, it is important to recall that Make defines dependencies by files. Therefore, the input and output of every step must be a file. Also recall that Make will use the modification date of the prerequisite and target files to see if the target must be re-built or not. Therefore during the processing, _many_ intermediate files will be created (see the tips section below on a good strategy to deal with large/huge files). To keep the source and (intermediate) built files separate, at configuration time, the user _must_ define a top-level build directory variable (or `$(BDIR)`) to host all the intermediate files. This directory doesn't need to be version controlled or even synchronized, or backed-up in other servers: its contents are all products of the pipeline, and can be easily re-created any time. As you define targets for your new rules, it is thus important to place them all under sub-directories of `$(BDIR)`. Let's start reviewing the processing with the top Makefile. Please open and inspect it as we go along here. The first step (un-commented line) defines the ultimate target (`paper.pdf`). You shouldn't modify this line. The rule to build `paper.pdf` is in another Makefile that will be imported into this top Makefile later. Don't forget that Make first scans the Makefile(s) once completely (to define dependencies and etc) and starts its execution after that. So it is fine to define the rule to build `paper.pdf` at a later stage (this is one beauty of Make!). Having defined the top target, our next step is to include all the other necessary Makefiles. First we include all Makefiles that satisfy this wildcard: `reproduce/config/pipeline/*.mk`. These Makefiles don't actually have any rules, they just have values for various free parameters throughout the pipeline. Open a few of them to see for your self. These Makefiles must only contain raw Make variables (pipeline configurations). By raw we mean that the Make variables in these files must not depend on variables in any other Makefile. This is because we don't want to assume any order in reading them. It is very important to *not* define any rule or other Make construct in any of these _configuration-Makefiles_ (see the next paragraph for Makefiles with rules). This will enable you to set the respective Makefiles in this directory as a prerequisite to any target that depends on their variable values. Therefore, if you change any of their values, all targets that depend on those values will be re-built. Once all the raw variables have been imported into the top Makefile, we are ready to import the Makefiles containing the details of the processing steps (Makefiles containing rules, let's call these _workhorse-Makefiles_). But in this phase *order is important*, because the prerequisites of most rules will be other rules that will be defined at a lower level (not a fixed name like `paper.pdf`). The lower-level rules must be imported into Make before the higher-level ones. Hence, we can't use a simple wildcard like when we imported configuration-Makefiles above. All these Makefiles are defined in `reproduce/src/make`, therefore, the top Makefile uses the `foreach` function to read them in a specific order. The main body of this pipeline is thus going to be managed within the workhorse-Makefiles that are in `reproduce/src/make`. If you set clear-to-understand names for these workhorse-Makefiles and follow the convention of the top Makefile that you only include one workhorse-Makefile per line, the `foreach` loop of the top Makefile that imports them will become very easy to read and understand by eye. This will let you know generally which step you are taking before or after another. Projects will scale up very fast. Thus if you don't start and continue with a clean and robust convention like this, in the end it will become very dirty and hard to manage/understand (even for yourself). As a general rule of thumb, break your rules into as many logically-similar but independent steps as possible. All processing steps are assumed to ultimately (usually after many rules) end up in some number, image, figure, or table that are to be included in the paper. The writing of the values into the final report is managed through separate LaTeX files that only contain macros (a name given to a number/string to be used in the LaTEX source, which will be replaced when compiling it to the final PDF). So usually the last target in a Makefile is a `.tex` file (with the same base-name as the Makefile, but in `$(BDIR)/tex/macros`). This intermediate TeX file rule will only contain commands to fill the TeX file up with values/names that were done in that Makefile. As a result, if the targets in a workhorse-Makefile aren't directly a prerequisite of other workhorse-Makefile targets, they should be a pre-requisite of that intermediate LaTeX macro file. In `reproduce/src/make/paper.mk` contains the rule to build `paper.pdf` (final target of the whole reproduction pipeline). If look in it, you will notice that it depends on `tex/pipeline.tex`. Therefore, last part of the top-level `Makefile` is the rule to build `tex/pipeline.tex`. `tex/pipeline.tex` is the connection between the processing steps of the pipeline, and the creation of the final PDF. Therefore, to keep the over-all management clean, the rule to create this bridge between the two phases is defined in the top-level `Makefile`. As you see in the top-level `Makefile`, `tex/pipeline.tex` is only a merging/concatenation of LaTeX macros defined as the output of each high-level processing step (the separate work-horse Makefiles that you included). One of the LaTeX macros created by `reproduce/src/make/initialize.mk` is `\bdir`. It is the location of the build directory. In some cases you want tables and images to also be included in the final PDF. To keep these necessary LaTeX inputs, you can define other directories under `$(BDIR)/tex` in the relevant workhorse-Makefile. You can then easily guide LaTeX to look into the proper directory to import an image for example through the `\bdir` macro. During the research, it often happens that you want to test a step that is not a prerequisite of any higher-level operation. In such cases, you can (temporarily) define the target of that rule as a prerequisite of `tex/pipeline.tex`. If your test gives a promising result and you want to include it in your research, set it as prerequisites to other rules and remove it from the list of prerequisites for `tex/pipeline.tex`. In fact, this is how a project is designed to grow in this framework. Summary ------- Based on the explanation above, some major design points you should have in mind are listed below. - Define new `reproduce/src/make/XXXXXX.mk` workhorse-Makefile(s) with good and human-friendly name(s) replacing `XXXXXX`. - Add `XXXXXX`, as a new line, to the loop which includes the workhorse-Makefiles in the top-level `Makefile`. - Do not use any constant numbers (or important names like filter names) in the workhorse-Makefiles or paper's LaTeX source. Define such constants as logically-grouped, separate configuration-Makefiles in `reproduce/config/pipeline`. Then set the respective configuration-Makefiles file as a pre-requisite to any rule that uses the variable defined in it. - Through any number of intermediate prerequisites, all processing steps should end in (be a prerequisite of) `tex/pipeline.tex`. `tex/pipeline.tex` is the bridge between the processing steps and PDF-building steps. Checklist to customize the pipeline =================================== Take the following steps to fully customize this pipeline for your research project. After finishing the list, be sure to run `./configure` and `make` to see if everything works correctly before expanding it. If you notice anything missing or any in-correct part (probably a change that has not been explained here), please let us know to correct it. As described above, the concept of a reproduction pipeline heavily relies on [version control](https://en.wikipedia.org/wiki/Version_control). Currently this pipline uses Git as its main version control system. If you are not already familiar with Git, please read the first three chapters of the [ProGit book](https://git-scm.com/book/en/v2) which provides a wonderful practical understanding of the basics. You can read later chapters as you get more advanced in later stages of your work. - **Get this repository and its history** (if you don't already have it): Arguably the easiest way to start is to clone this repository as shown below. The main branch of this pipeline is called `pipeline`. This allows you to use the common branch name `master` for your own research, while keeping up to date with improvements in the pipeline. ```shell $ git clone https://gitlab.com/makhlaghi/reproducible-paper.git $ mv reproducible-paper my-project-name # Your own directory name. $ cd my-project-name # Go into the cloned directory. $ git remote rename origin pipeline-origin # Rename the pipeline's remote. $ git checkout -b master # Create, enter master branch. ``` - **Test the pipeline**: Before making any changes, it is important to test the pipeline and see if everything works properly with the commands below. If there is any problem in the `./configure` or `make` steps, please contact us to fix the problem before continuing. Since the building of dependencies in `./configure` can take long, you can take the next few steps (editing the files) while its working (they don't affect the configuration). After `make` is finished, open `paper.pdf` and if it looks fine, you are ready to start customizing the pipeline for your project. But before that, clean all the extra pipeline outputs with `make clean` as shown below. ```shell $ ./configure # Set top directories and build dependencies. $ make # Run the pipeline. # Open 'paper.pdf' and see if everything is ok. $ make clean # Delete high-level outputs. ``` - **Copyright**, **name** and **date**: Go over the existing scripting files and add your name and email to the copyright notice. You can find the files by searching for the placeholder email `your@email.address` (which you should change) with the command below (you can ignore this file, `README.md`, and any in the `tex/` directory). Don't forget to add your name after the copyright year also. When making new files, always remember to add a similar copyright statement at the top of the file and also ask your colleagues to do so when they edit a file. This is very important. ```shell $ grep -r your@email.address ./* ``` - **Title**, **short description** and **author** in source files: In this raw skeleton, the title or short description of your project should be added in the following two files: `reproduce/src/make/Top-Makefile` (the first line), and `tex/preamble-header.tex`. In both cases, the texts you should replace are all in capital letters to make them easier to identify. Of course, if you use a different LaTeX method of managing the title and authors, please feel free to use your own methods after finishing this checklist and doing your first commit. - **Gnuastro**: GNU Astronomy Utilities (Gnuastro) is currently a dependency of the pipeline which will be built and used. The main reason for this is to demonstrate how critically important it is to version your software. If you will be using Gnuastro, you can also remove the "marks" (comments) put in the relevant files of the list below to make them more readable. - Delete marked part(s) in `configure`. - Delete `astnoisechisel` from the value of `top-level-programs` in `reproduce/src/make/dependencies.mk`. You can keep the rule to build `astnoisechisel`, since its not in the `top-level-programs` list, it (and all its dependencies) will be ignored. - Delete marked parts in `reproduce/src/make/initialize.mk`. - Delete `and Gnuastro \gnuastroversion` from `tex/preamble-style.tex`. - **Other dependencies**: If there are any more of the dependencies that you don't use (or others that you need), then remove (or add) them in the respective parts of `reproduce/src/make/dependencies.mk`. It is commented thoroughly and reading over the comments should guide you on what to add/remove and where. - **`README`**: Go through this top-level instruction file and make it fit to your pipeline: update the text and etc. Don't forget that your colleagues or anyone else, will first be drawn to read this file, so make it as easy as possible for them to understand your work. Therefore, also check and update `README` one last time when you are ready to publish your work (and its reproduction pipeline). - **Input dataset (can be done later)**: The user manages the top-level directory of the input data through the variables set in `reproduce/config/pipeline/LOCAL.mk.in` (the user actually edits a `LOCAL.mk` file that is created by `configure` from the `.mk.in` file, but the `.mk` file is not under version control). Datasets are usually large and the users might already have their copy don't need to download them). So you can define a variable (all in capital letters) in `reproduce/config/pipeline/LOCAL.mk.in`. For example if you are working on data from the XDF survey, use `XDF`. You can use this variable to identify the location of the raw inputs on the running system. Here, we'll assume its name is `SURVEY`. Afterwards, change any occurrence of `SURVEY` in the whole pipeline with the new name. You can find the occurrences with a simple command like the ones shown below. We follow the Make convention here that all `ONLY-CAPITAL` variables are those directly set by the user and all `small-caps` variables are set by the pipeline designer. All variables that also depend on this survey have a `survey` in their name. Hence, also correct all these occurrences to your new name in small-caps. Of course, ignore/delete those occurrences that are irrelevant, like those in this file. Note that in the raw version of this template no target depends on these files, so they are ignored. Afterwards, set the webpage and correct the filenames in `reproduce/src/make/download.mk` if necessary. ```shell $ grep -r SURVEY ./ $ grep -r survey ./ ``` - **Other input datasets (can be done later)**: Add any other input datasets that may be necessary for your research to the pipeline based on the example above. - **Delete dummy parts (can be done later)**: The template pipeline contains some parts that are only for the initial/test run, not for any real analysis. The respective files to remove and parts to fix are discussed here. - `paper.tex`: Delete the text of the abstract and the paper's main body, *except* the "Acknowledgements" section. This reproduction pipeline was designed by funding from many grants, so its necessary to acknowledge them in your final research. - `Makefile`: Delete the two lines containing `delete-me` in the `foreach` loops. Just make sure the other lines that end in `\` are immediately after each other. - Delete all `delete-me*` files in the following directories: ```shell $ rm tex/delete-me* $ rm reproduce/src/make/delete-me* $ rm reproduce/config/pipeline/delete-me* ``` - **Your first commit**: You have already made some small and basic changes in the steps above and you are in the `master` branch. So, you can officially make your first commit in your project's history. But before that you need to make sure that there are no problems in the pipeline (this is a good habit to always re-build the system before a commit to be sure it works as expected). ```shell $ ./.local/bin/make clean # Delete outputs ('make distclean' for everything) $ ./.local/bin/make # Build the pipeline to ensure everything is fine. $ git add -u # Stage all the changes. $ git status # Make sure everything is fine. $ git commit # Your first commit, add a nice description. $ git tag -a v0 # Tag this as the zero-th version of your pipeline. ``` - **Setup the remote**: You can use any [hosting facility](https://en.wikipedia.org/wiki/Comparison_of_source_code_hosting_facilities) that supports Git to keep an online copy of your project's version controlled history. We recommend [GitLab](https://gitlab.com) because it allows any number of private repositories for free and because you can host GitLab on your own server. Create an account in your favorite hosting facility (if you don't already have one), and define a new project there. It will give you a link (ending in `.git`) that you can put in place of `XXX` in the command below. ```shell git remote add origin XXX git push -u origin --all git push -u origin --tags ``` - **Start your exciting research**: You are now ready to add flesh and blood to this raw skeleton by further modifying and adding your exciting research steps. You can use the "published works" section in the introduction (above) as some fully working models to learn from. Also, don't hesitate to contact us if you have any questions. Any time you are ready to push your commits to the remote repository, you can simply use `git push`. - **Pre-publication: add notice on reproducibility**: Add a notice somewhere prominent in the first page within your paper, informing the reader that your research is fully reproducible. For example in the end of the abstract, or under the keywords with a title like "reproducible paper". This will encourage them to publish their own works in this manner also and also will help spread the word. - **Feedback**: As you use the pipeline you will notice many things that if implemented from the start would have been very useful for your work. This can be in the actual scripting and architecture of the pipeline or in useful implementation and usage tips, like those below. In any case, please share your thoughts and suggestions with us, so we can add them here for everyone's benefit. Usage tips: designing your pipeline/workflow ============================================ The following is a list of design points, tips, or recommendations that have been learned after some experience with this pipeline. Please don't hesitate to share any experience you gain after using this pipeline with us. In this way, we can add it here for the benefit of others. - **Modularity**: Modularity is the key to easy and clean growth of a project. So it is always best to break up a job into as many sub-components as reasonable. Here are some tips to stay modular. - *Short recipes*: if you see the recipe of a rule becoming more than a handful of lines which involve significant processing, it is probably a good sign that you should break up the rule into its main components. Try to only have one major processing step per rule. - *Context-based (many) Makefiles*: This pipeline is designed to allow the easy inclusion of many Makefiles (in `reproduce/src/make/*.mk`) for maximal modularity. So keep the rules for closely related parts of the processing in separate Makefiles. - *Descriptive names*: Be very clear and descriptive with the naming of the files and the variables because a few months after the processing, it will be very hard to remember what each one was for. Also this helps others (your collaborators or other people reading the pipeline after it is published) to more easily understand your work and find their way around. - *Naming convention*: As the project grows, following a single standard or convention in naming the files is very useful. Try best to use multiple word filenames for anything that is non-trivial (separating the words with a `-`). For example if you have a Makefile for creating a catalog and another two for processing it under models A and B, you can name them like this: `catalog-create.mk`, `catalog-model-a.mk` and `catalog-model-b.mk`. In this way, when listing the contents of `reproduce/src/make` to see all the Makefiles, those related to the catalog will all be close to each other and thus easily found. This also helps in auto-completions by the shell or text editors like Emacs. - *Source directories*: If you need to add files in other languages for example in shell, Python, AWK or C, keep them in a separate directory under `reproduce/src`, with the appropriate name. - *Configuration files*: If your research uses special programs as part of the processing, put all their configuration files in a devoted directory (with the program's name) within `reproduce/config`. Similar to the `reproduce/config/gnuastro` directory (which is put in the template as a demo in case you use GNU Astronomy Utilities). It is much cleaner and readable (thus less buggy) to avoid mixing the configuration files, even if there is no technical necessity. - **Contents**: It is good practice to follow the following recommendations on the contents of your files, whether they are source code for a program, Makefiles, scripts or configuration files (copyrights aren't necessary for the latter). - *Copyright*: Always start a file containing programming constructs with a copyright statement like the ones that this template starts with (for example in the top level `Makefile`). - *Comments*: Comments are vital for readability (by yourself in two months, or others). Describe everything you can about why you are doing something, how you are doing it, and what you expect the result to be. Write the comments as if it was what you would say to describe the variable, recipe or rule to a friend sitting beside you. When writing the pipeline it is very tempting to just steam ahead with commands and codes, but be patient and write comments before the rules or recipes. This will also allow you to think more about what you should be doing. Also, in several months when you come back to the code, you will appreciate the effort of writing them. Just don't forget to also read and update the comment first if you later want to make changes to the code (variable, recipe or rule). As a general rule of thumb: first the comments, then the code. - *File title*: In general, it is good practice to start all files with a single line description of what that particular file does. If further information about the totality of the file is necessary, add it after a blank line. This will help a fast inspection where you don't care about the details, but just want to remember/see what that file is (generally) for. This information must of course be commented (its for a human), but this is kept separate from the general recommendation on comments, because this is a comment for the whole file, not each step within it. - **Make programming**: Here are some experiences that we have come to learn over the years in using Make and are useful/handy in research contexts. - *Automatic variables*: These are wonderful and very useful Make constructs that greatly shrink the text, while helping in read-ability, robustness (less bugs in typos for example) and generalization. For example even when a rule only has one target or one prerequisite, always use `$@` instead of the target's name, `$<` instead of the first prerequisite, `$^` instead of the full list of prerequisites and etc. You can see the full list of automatic variables [here](https://www.gnu.org/software/make/manual/html_node/Automatic-Variables.html). If you use GNU Make, you can also see this page on your command-line: ```shell $ info make "automatic variables ``` - *Debug*: Since Make doesn't follow the common top-down paradigm, it can be a little hard to get accustomed to why you get an error or un-expected behavior. In such cases, run Make with the `-d` option. With this option, Make prints a full list of exactly which prerequisites are being checked for which targets. Looking (patiently) through this output and searching for the faulty file/step will clearly show you any mistake you might have made in defining the targets or prerequisites. - *Large files*: If you are dealing with very large files (thus having multiple copies of them for intermediate steps is not possible), one solution is the following strategy. Set a small plain text file as the actual target and delete the large file when it is no longer needed by the pipeline (in the last rule that needs it). Below is a simple demonstration of doing this, where we use Gnuastro's Arithmetic program to add all pixels of the input image with 2 and create `large1.fits`. We then subtract 2 from `large1.fits` to create `large2.fits` and delete `large1.fits` in the same rule (when its no longer needed). We can later do the same with `large2.fits` when it is no longer needed and so on. ``` large1.fits.txt: input.fits astarithmetic $< 2 + --output=$(subst .txt,,$@) echo "done" > $@ large2.fits.txt: large1.fits.txt astarithmetic $(subst .txt,,$<) 2 - --output=$(subst .txt,,$@) rm $(subst .txt,,$<) echo "done" > $@ ``` A more advanced Make programmer will use Make's [call function](https://www.gnu.org/software/make/manual/html_node/Call-Function.html) to define a wrapper in `reproduce/src/make/initialize.mk`. This wrapper will replace `$(subst .txt,,XXXXX)`. Therefore, it will be possible to greatly simplify this repetitive statement and make the code even more readable throughout the whole pipeline. - **Dependencies**: It is critically important to exactly document, keep and check the versions of the programs you are using in the pipeline. - *Check versions*: In `reproduce/src/make/initialize.mk`, check the versions of the programs you are using. - *Keep the source tarball of dependencies*: keep a tarball of the necessary version of all your dependencies (and also a copy of the higher-level libraries they depend on). Software evolves very fast and only in a few years, a feature might be changed or removed from the mainstream version or the software server might go down. To be safe, keep a copy of the tarballs. Software tarballs are rarely over a few megabytes, very insignificant compared to the data. If you intend to release the pipeline in a place like Zenodo, then you can create your submission early (before public release) and upload/keep all the necessary tarballs (and data) there. [zenodo.1163746](https://doi.org/10.5281/zenodo.1163746) is one example of how the data, Gnuastro (main software used) and all major Gnuastro's dependencies have been uploaded with the pipeline. - *Keep your input data*: The input data is also critical to the pipeline, so like the above for software, make sure you have a backup of them. - **Version control**: It is important (and extremely useful) to have the history of your pipeline under version control. So try to make commits regularly (after any meaningful change/step/result), while not forgetting the following notes. - *Tags*: To help manage the history, tag all major commits. This helps make a more human-friendly output of `git describe`: for example `v1-4-gaafdb04` states that we are on commit `aafdb04` which is 4 commits after tag `v1`. The output of `git describe` is included in your final PDF as part of this pipeline. Also, if you use reproducibility-friendly software like Gnuastro, this value will also be included in all output files, see the description of `COMMIT` in [Output headers](https://www.gnu.org/software/gnuastro/manual/html_node/Output-headers.html). In the checklist above, we tagged the first commit of your pipeline with `v0`. Here is one suggestion on when to tag: when you have fully adopted the pipeline and have got the first (initial) results, you can make a `v1` tag. Subsequently when you first start reporting the results to your colleagues, you can tag the commit as `v2`. Afterwards when you submit to a paper, it can be tagged `v3` and so on. - *Pipeline outputs*: During your research, it is possible to checkout a specific commit and reproduce its results. However, the processing can be time consuming. Therefore, it is useful to also keep track of the final outputs of your pipeline (at minimum, the paper's PDF) in important points of history. However, keeping a snapshot of these (most probably large volume) outputs in the main history of the pipeline can unreasonably bloat it. It is thus recommended to make a separate Git repo to keep those files and keep this pipeline's volume as small as possible. For example if your main pipeline is called `my-exciting-project`, the name of the outputs pipeline can be `my-exciting-project-output`. This enables easy sharing of the output files with your co-authors (with necessary permissions) and not having to bloat your email archive with extra attachments (you can just share the link to the online repo in your communications). After the research is published, you can also release the outputs pipeline, or you can just delete it if it is too large or un-necessary (it was just for convenience, and fully reproducible after all). This pipeline's output is available for demonstration in the separate [reproducible-paper-output](https://gitlab.com/makhlaghi/reproducible-paper-output) repository. Future improvements =================== This is an evolving project and as time goes on, it will evolve and become more robust. Here are the list of features that we plan to add in the future. - *Containers*: It is important to have better/full control of the environment of the reproduction pipeline. Our current reproducible paper pipeline builds the higher-level programs (for example GNU Bash, GNU Make, GNU AWK and etc) it needs and sets `PATH` to prefer its own builds. It currently doesn't build and use its own version of lower-level tools (like the C library and compiler). We plan to add the build steps of these low level tools so the system's `PATH' can be completely ignored within the pipeline and we are in full control of the whole build process. Another solution is based on [an interesting tutorial](https://mozillafoundation.github.io/2017-fellows-sf/re-papers/index.html) by the Mozilla science lab to build reproducible papers. It suggests using the [Nix package manager](https://nixos.org/nix/about.html) to build the necessary software for the pipeline and run the pipeline in its completely closed environment. This is an interesting solution because using Nix or [Guix](https://www.gnu.org/software/guix/) (which is based on Nix, but uses the [Scheme language](https://en.wikipedia.org/wiki/Scheme_(programming_language)), not a custom language for the management) will allow a fully working closed environment on the host system which contains the instructions on how to build the environment. The availability of the instructions to build the programs and environment with Nix or Guix, makes them a better solution than binary containers like [docker](https://www.docker.com/) which are essentially just a binary (not human readable) black box and only usable on the given CPU architecture. However, one limitation of using these is their own installation (which usually requires root access). Appendix: Necessity of exact reproduction in scientific research ================================================================ In case [the link above](http://akhlaghi.org/reproducible-science.html) is not accessible at the time of reading, here is a copy of the introduction of that link, describing the necessity for a reproduction pipeline like this (copied on February 7th, 2018): The most important element of a "scientific" statement/result is the fact that others should be able to falsify it. The Tsunami of data that has engulfed astronomers in the last two decades, combined with faster processors and faster internet connections has made it much more easier to obtain a result. However, these factors have also increased the complexity of a scientific analysis, such that it is no longer possible to describe all the steps of an analysis in the published paper. Citing this difficulty, many authors suffice to describing the generalities of their analysis in their papers. However, It is impossible to falsify (or even study) a result if you can't exactly reproduce it. The complexity of modern science makes it vitally important to exactly reproduce the final result. Because even a small deviation can be due to many different parts of an analysis. Nature is already a black box which we are trying so hard to comprehend. Not letting other scientists see the exact steps taken to reach a result, or not allowing them to modify it (do experiments on it) is a self-imposed black box, which only exacerbates our ignorance. Other scientists should be able to reproduce, check and experiment on the results of anything that is to carry the "scientific" label. Any result that is not reproducible (due to incomplete information by the author) is not scientific: the readers have to have faith in the subjective experience of the authors in the very important choice of configuration values and order of operations: this is contrary to the scientific spirit.