On this page, we first describe how the files are distributed in the directory hierarchy. We then move on to show how to find your way around using simple data structures in the Project paths section, so that you just need to make changes in a single place (remember to minimise code repetition!).

Directory structure

The left node of the following graph shows the contents of the project root directory after executing python configure build install:


Files and directories in brownish colours are constructed by Waf; those with a bluish background are added directly by the researcher. You immediately see the separation of inputs and outputs (one of our guiding principles) at work:

  • All source code is in the src directory.

  • All outputs are constructed in the bld directory.

  • The other objects in square brackets are put there during Waf’s install phase, so that they can be opened easily (paper, presentation, documentation).

  • The remainder is made up of objects related to Waf:

    • is the file that starts up Waf (you will never need to change it).
    • wscript is the main entry point for the instructions we give to Waf.
    • .mywaflib contains Waf’s internals.

The contents of both the root/bld/out and the root/src directories directly follow the steps of the analysis from the workflow section (you can usually ignore the ro../bld directory, except when you need to take a look at LaTeX log-files).

The idea is that everything that needs to be run during the, say, analysis step, is specified in root/src/analysis and all its output is placed in root/bld/out/analysis.

Some differences:

  • Because they are accessed frequently, figures and tables get extra directories in root/bld/out next to final

  • The directory root/src contains many more subdirectories:

    • original_data is the place to store the data in its raw form, as downloaded / transcribed / … The original data should never be modified and saved under the same name.
    • model_code contains source files that might differ by model and that are potentially used at various steps of the analysis.
    • model_specs contains JSON files with model specifications. The choice of JSON is motivated by the attempt to be language-agnostic: JSON is quite expressive and there are parsers for nearly all languages (for Stata there is a converter in the wscript file of the Stata version of the template)
    • library provides code that may be used by different steps of the analysis. Little code snippets for input / output or stuff that is not directly related to the model would go here. The distinction from the model_code directory is a bit arbitrary, but I have found it useful in the past.

As an example of how things look further down in the hierarchy, consider the analysis step that was described here:


Remember that the script root/src/analysis/ is run with an argument baseline or max_moves_2. The code then accesses the respective file in root/src/model_specs, root/src/model_code/, and bld/out/data/initial_locations.csv (not shown). These are many different locations to keep track of; your project organisation will change as your project evolves and typing in entire paths at various locations is cumbersome. The next sections shows how this is solved in the project template.

Project paths

The first question to ask is whether we should be working with absolute or relative paths. Let us first consider the pros and cons of each.

  • Relative paths (e.g., ..\model_code\ or ../model_code/

    • Pro: Portable across machines; provide abstraction from irrelevant parts of underlying directory structure.
    • Con: Introduction of state (the directory used as starting point), which is bad for maintainability and reproducibility.
  • Absolute paths (e.g., C:\projects\schelling\src\model_code\ or /Users/xxx/projects/schelling/src/model_code/

    • Pro: Any file or directory is unambiguously specified.
    • Con: Not portable across machines.

The project template combines the best of both worlds by requiring you to specify relative paths for all often-accessed locations in the main wscript file. These are then used throughout the project template – both in the wscript files and in any substantial code. The next sections show how to specify them and how to use them in different circumstances.

Specifying project paths in the main wscript file

This is how the project paths are specified in the main wscript file:

top = '.'
out = 'bld'

def set_project_paths(ctx):
    """Return a dictionary with project paths represented by Waf nodes."""

    pp = {}
    pp['PROJECT_ROOT'] = '.'
    pp['IN_DATA'] = 'src/original_data'
    pp['IN_MODEL_CODE'] = 'src/model_code'
    pp['IN_MODEL_SPECS'] = 'src/model_specs'
    pp['OUT_DATA'] = '{}/out/data'.format(out)
    pp['OUT_ANALYSIS'] = '{}/out/analysis'.format(out)
    pp['OUT_FINAL'] = '{}/out/final'.format(out)
    pp['OUT_FIGURES'] = '{}/out/figures'.format(out)
    pp['OUT_TABLES'] = '{}/out/tables'.format(out)

    # Convert the directories into Waf nodes.
    for key, val in pp.items():
        pp[key] = ctx.path.make_node(val)

    return pp

All these paths are relative to the project root, so you can directly use them on many different machines. Note the distinction between IN and OUT in the keys and that we prefix all of the latter by bld.

The mappings from input to output by step of the analysis should be easy enough from the names:

  1. data_management, original_dataOUT_DATA
  2. analysisOUT_ANALYSIS

In addition, there are the “special” input directories library, model_code, and model_specs, of course.

Usage of the project paths within wscript files

The first thing to do is to make these project paths available in wscript files further down the directory hierarchy. We do so in the build function of root/wscript; the relevant lines are:

def build(ctx):
    ctx.env.PROJECT_PATHS = set_project_paths(ctx)
    ctx.path_to = path_to

The first line of the function attaches the project paths we defined in the previous section to the build context object. The second attaches a convenience function to the same object, which will do all the heavy lifting. You do not need to care about its internals, only about its interface:

ctx.path_to(ctx, pp_key, *args)

Return the relative path to os.path.join(args) in the directory PROJECT_PATHS[pp_key] as seen from ctx.path (i.e. the directory of the current wscript).

Use this to get the relative path—as needed by Waf—to a file in one of the directory trees defined in the PROJECT_PATHS dictionary above.

This description may be a bit cryptic, but it says it all: Waf needs paths relative to the wscript where you define a task generator. This function returns it. You always need to supply three arguments:

  1. The build context (completely mechanical, always the same)
  2. The key of the directory you want to access.
  3. The name of the file in the directory. If there is a further hierarchy of directories, separate directory and file names by commas.

Let us look at root/src/analysis/wscript as an example again:

#! python

def build(ctx):

    for model in 'baseline', 'max_moves_2':

        # Illustrate use of run_py_script with automatic model specification.
                ctx.path_to(ctx, 'OUT_DATA', 'initial_locations.csv'),
                ctx.path_to(ctx, 'IN_MODEL_CODE', ''),
                ctx.path_to(ctx, 'IN_MODEL_SPECS', '{}.json'.format(model)),
                ctx.path_to(ctx, 'OUT_ANALYSIS', 'schelling_{}.pickle'.format(model)),
                ctx.path_to(ctx, 'OUT_ANALYSIS', 'log', 'schelling_{}.log'.format(model))

Note that the order of the arguments is the same in each of the five calls of ctx.path_to(). The last one has an example of a nested directory structure: We do not need the log-files very often and they only clutter up the OUT_ANALYSIS directory, so we put them in a subdirectory.

Usage of the project paths in substantial code

The first thing to do is to specify a task generator that writes a header with project paths to disk. This is done using the write_project_paths feature. The following line is taken from the build function in root/wscript:

    # Generate header file with project paths in 'bld' directory
    ctx(features='write_project_paths', target='')

The write_project_paths feature is smart: It will recognise the syntax for its target by the extension you add to the latter. Currently supported: .py, .do, .m, .r, .pm.

The paths contained in the resulting file (root/bld/ are absolute paths, so you do not need to worry about the location of your interpreter etc.

The exact usage varies a little bit by language; see the respective template for examples. In Python, you first import a function called project_paths_join:

from bld.project_paths import project_paths_join as ppj

You can then use it to obtain absolute paths to any location within your project. E.g., for the log-file in the analysis step, you would use:

ppj("OUT_ANALYSIS", "log", "schelling_{}.log".format(model_name))

When you need to change the paths for whatever reason, you just need to updated them once in the main wscript file; everything else will work automatically. Even if you need to change the keys – e.g. because you want to break the analysis step into two – you can easiliy search and replace OUT_ANALYSIS in the entire project.