j2gpp 2.2.1

A Jinja2-based General Purpose Preprocessor

J2GPP - Jinja2-based General Purpose Preprocessor

j2gpp is a command-line tool for rendering templates using the Jinja2 syntax. It's intended purpose is to serve as a preprocessor for any programming or markup language with a unified syntax and flow across languages.

• J2GPP - Jinja2-based General Purpose Preprocessor
  • Installation
  • Basic usage
  • Command line arguments
    • Specify output directory
    • Specifying output file
    • Include search directory
    • Passing global variables in command line
    • Loading global variables from files
    • Loading global variables from environment
    • Loading custom Jinja2 filters
    • Loading custom Jinja2 tests
    • Processing variables before rendering
    • Option flags
    • Context variables
    • Built-in filters
  • Process directories
  • Supported formats for variables
    • Command line define
    • YAML
    • JSON
    • XML
    • TOML
    • INI/CFG
    • ENV
    • CSV/TSV
  • Scripting in J2GPP templates
    • Conditional and do extension
    • Conditional and filters
    • Throw errors and warnings from template
    • Writing files
      • Write example
      • Append example
      • Path argument
      • Writing to both files
      • Skipping the parent file

Installation

With Python >= 3.7, Install from Pypi :

pip install j2gpp

Basic usage

j2gpp requires at least one source be provided. The source paths can be files or directories, relative or absolute, and can use UNIX-style patterns such as wildcards. Template file names must end with the .j2 extension which will be stripped at render.

For more information about the Jinja2 syntax, see the documentation at jinja.palletsprojects.com.

For instance, suppose we have a templatized source file foo.c.j2 :

#include <stdio.h>

{% set message = "Hello, world!" %}

int main() {
  printf("{{message}}");
  return 0;
}

We can render the template using j2gpp :

j2gpp ./foo.c.j2

The output is written to foo.c next to the source template :

#include <stdio.h>

int main() {
  printf("Hello, world!");
  return 0;
}

The following arguments can be added to the command for additional features. The details of each command is explained in the sections below.

Argument	Description
-O/--outdir	Output directory for all rendered templates
-o/--output	Output file for single template
-I/--incdir	Include search directory for include and import statements
-D/--define	Inline global variables for all templates
-V/--varfile	Global variables files for all templates
--envvar	Loads environment variables as global variables
--filters	Load extra Jinja2 filters from a Python file
--tests	Load extra Jinja2 tests from a Python file
--file-vars-adapter	Load a Python function to process variables after loading from a file
--global-vars-adapter	Load a Python function to process all variables before rendering
--overwrite-outdir	Overwrite output directory
--warn-overwrite	Warn when overwriting files
--no-overwrite	Prevent overwriting files
--no-strict-undefined	Disable error with undefined variable in template
--no-check-identifier	Disable warning when attributes are not valid identifiers
--fix-identifiers	Replace invalid characters from identifiers with underscore
--csv-delimiter	CSV delimiter (default: ',')
--csv-escapechar	CSV escape character (default: None)
--csv-dontstrip	Disable stripping whitespace of CSV values
--render-non-template	Process also source files that are not recognized as templates
--copy-non-template	Copy source files that are not templates to output directory
--force-glob	Glob UNIX-like patterns in path even when quoted
--perf	Measure the execution time for performance testing
--version	Print J2GPP version and quits
--license	Print J2GPP license and quits

Command line arguments

Specify output directory

By default the rendered files are saved next to the source templates. You can provide an output directory with the -O/--outdir argument. The output directory path can be relative or absolute. If the directory doesn't exist, it will be created.

For instance the following command will write the rendered file to ./bar/foo.c.

j2gpp ./foo.c.j2 --outdir ./bar/

Specifying output file

By default the rendered files are saved next to the source templates. If a single source template is provided, you can specify the output file directory and name with the -o/--output argument. The output file path can be relative or absolute. If the directory doesn't exist, it will be created.

For instance the following command will write the rendered file to ./bar.c.

j2gpp ./foo.c.j2 --output ./bar.c

Include search directory

The include and import Jinja2 statements require specifying the directory in which the renderer will search. That is provided using the -I/--incidr argument.

For instance, with the following command, the files in the directory ./includes/ will be available to include and import statements when rendering the template foo.c.j2.

j2gpp ./foo.c.j2 --incdir ./includes/

Passing global variables in command line

You can pass global variables to all templates rendered using the -D/--define argument with a list of variables in the format name=value. Values are parsed to cast to the correct Python type as explained later. Dictionary attributes to any depth can be assigned using dots "." to separate the keys. Global variables defined in the command line overwrite the global variables set by loading files.

For instance, with the following command, the variable bar will have the value 42 when rendering the template foo.c.j2.

j2gpp ./foo.c.j2 --define bar=42

Loading global variables from files

You can load global variables from files using the -V/--varfile argument with a list of files. The file paths can be relative or absolute, and can use UNIX-style patterns such as wildcards. Variables file types supported right now are YAML, JSON, XML, TOML, INI/CFG, ENV, CSV and TSV. Global variables loaded from files are overwritten by variables defined in the command line.

For instance, with the following command, the variable bar will have the value 42 when rendering the template foo.c.j2.

j2gpp ./foo.c.j2 --varfile ./qux.yml

With the variables file qux.yml :

bar: 42

Loading global variables from environment

You can import the environment variables of the shell as global variables using the --envvar argument. The name of the variables will be that of the environment variable and the value will be cast automatically to the proper Python/Jinja2 type.

For instance, with the following command, the variable BAR will have the value 42 when rendering the template foo.c.j2.

export BAR=42
j2gpp ./foo.c.j2 --envvar

If a string is provided after the --envvar argument, the environment variables will be stored in an object of the name provided instead of at the root.

For instance, with the following command, the variable ENV.BAR will have the value 42 when rendering the template foo.c.j2.

export BAR=42
j2gpp ./foo.c.j2 --envvar ENV

Loading custom Jinja2 filters

You can import custom Jinja2 filters by providing Python files with the --filters argument. All functions defined in the python files will be available as Jinja2 filters in the templates.

For instance, with the following command and python file, the filter right_ajust will be available when rendering the template foo.c.j2.

j2gpp ./foo.c.j2 --filters ./bar.py

# bar.py
def right_ajust(s, length=0):
  return s.rjust(length)

Loading custom Jinja2 tests

You can import custom Jinja2 tests by providing Python files with the --tests argument. All functions defined in the python files will be available as Jinja2 tests in the templates.

For instance, with the following command and python file, the test prime will be available when rendering the template foo.c.j2.

j2gpp ./foo.c.j2 --tests ./bar.py

# bar.py
import math
def prime(x):
  if x<=1: return False
  for i in range(2,int(math.sqrt(x))+1):
    if (x%i) == 0:
      return False
  return True

Processing variables before rendering

You can perform transformations on the dictionary storing the variables before rendering templates by providing a Python function with the argument --global-vars-adapter. It takes two arguments, the first is the path to the Python script, and the second is the name of the function to use. The function must take in a single argument, the variables dictionary, and modify it as a reference (not return the modified dictionary). You can also provide an adapter function to be run on the variables loaded from each file before they are written to the global variables object by providing a function with the argument --file-vars-adapter.

For instance, with the following command and python file, the variable dictionary loaded from the file qux.yml will be processed by the function shout_values() before rendering the template foo.c.j2.

j2gpp ./foo.c.j2 --varfile ./qux.yml --file-vars-adapter ./bar.py shout_values

# bar.py
def shout_values(var_dict):
  for key,val in var_dict.items():
    if isinstance(val, str):
      var_dict[key] = val.upper()

Option flags

Some arguments are flags to enable or disable special features. This is more advanced but can be useful in niche situations.

--overwrite-outdir cleans the output directory before rendering the templates and copying any other files.

--warn-overwrite enables warnings triggered whenever a file is overwritten.

--no-overwrite prevents any file from being overwritten, and triggers a warning when that happens.

--no-strict-undefined disables errors triggered whenever a template variable is used but not defined.

--no-check-identifier disables the ckecking that the variables names and attributes are valid Python identifiers. Root variables with a name not passing this check will not be accessible in Jinja2 templates.

--fix-identifiers fixes variables and attributes names that are not valid Python identifiers by replacing incorrect characters by underscores, and if the first character is a number, an underscore is added before.

--csv-delimiter followed by a string will change the delimiter used to parse CSV variables files. The default is ",".

--csv-escapechar followed by a character sets the escape character used to parse CSV variables files. There is no escape character by default.

--csv-dontstrip disables the stripping of whitespace from CSV keys and values.

--render-non-template forces every source file found to be rendered, even if they are not recognized as a template (by ending with a template extension). The resulting file will be saved in the location following the rules of regular templates, but instead of removing the template extension, they will have a suffix added before the file extensions. By default, this suffix is _j2gpp, but this can be replaced by whatever is specified after the flag argument.

--copy-non-template enables the copying of the source files that are not recognized as templates or the files in the source directories to the output directory when one is provided with the --outdir argument.

--force-glob enables globbing UNIX-like patterns in the source files paths even if they are surrounded by quotes. This is disabled by default to allow processing files with * and [...] in their path. Paths provided without quotes are preprocessed by the shell and any wildcard or other patterns cannot be prevented.

Context variables

Useful context variables are added before any other variable is loaded. Some are global for all templates rendered, and some are template-specific.

Variable	Scope	Description
__python_version__	Global	Python version
__jinja2_version__	Global	Jinja2 version
__j2gpp_version__	Global	J2GPP version
__user__	Global	Name of the current user
__pid__	Global	Process ID of the current process
__ppid__	Global	Process ID of the parent process
__working_directory__	Global	Working directory
__output_directory__	Global	Output directory
__date__	Global	Date in the format DD-MM-YYYY
__date_inv__	Global	Date in the format YYYY-MM-DD
__time__	Global	Time in the format hh:mm:ss
__datetime__	Global	Timestamp in the format YYYY-MM-DD hh:mm:ss
__source_path__	Template	Path of the source template file
__output_path__	Template	Path where the template is rendered

Built-in filters

In addition to the Jinja2 built-in filters, J2GPP also defines many useful filter functions.

All functions from the Python libraries math and statistics are made available as filters. This includes useful functions such as sqrt, pow, log, sin, cos, floor, ceil, mean, median, variance, stdev, ...

An operation can be applied to all elements of a list with the filters list_add, list_sub, list_mult, list_div, list_mod, list_rem and list_exp respectively for the Python operators +, -, *, /, %, // and **.

Text alignment can be controlled with the Python functions ljust, rjust and center.

Case and formatting can be controlled with the Python functions title and swapcase, or the added functions camel, pascal, snake and kebab. camel and pascal will remove the underscores and hyphens by default but leave the dots ; that behaviour can be changed by providing the filter arguments remove_underscore, remove_hyphen and remove_dot as True or False. snake and kebab will group capitalized letters and preserve capitalization by default ; that behaviour can be changed by providing the filter arguments preserve_caps and group_caps as True or False.

Paragraph formatting is facilitated by multiple filters that should be used on a filter block. reindent removes pre-existing indentation and sets new one. autoindent removes pre-existing indentation and sets new indent by incrementing and decrementing the depth based on start and end delimiters of blocks provided by the starts and ends lists of strings provided by argument (culry braces by default). align aligns every line of the paragraph by columns, left before §, right before §§.

When using a list of dictionaries with a key in common, you can get the list element with the minimum or maximum value of that attribute using the filters el_of_max_attr or el_of_min_attr.

When using two-level dictionaries, the key corresponding to the minimum or maximum with regards to a specified attribute using the filters key_of_max_attr or key_of_min_attr.

You can count the number of occurrences of a value in a list using the count filter.

The write and append filters can be used to export the content of a filter to another file whose path is provided as argument to the filter. The path can be absolute or relative to the output rendered base template. By default, the content of the filter is not written to the base rendered template ; this behaviour can be changed by providing the filter argument preserve as True. The source template can also be prevented from resulting in a generated file by providing the filter argument write_source as False, and only the content of write and append blocks will generate files.

The warning and error filters can be used to throw warnings and errors from the template that will be displayed in the J2GPP logs. The filter is applied to a block, replaces the block with nothing and throws the warning or error with the content of the block as comment. The filter works with conditional blocks if the version of Jinja2 installed supports the @render_time_only decorator.

Process directories

When the source path provided corresponds to a directory, J2GPP will look for any template files in the source directory tree. If no output directory argument is provided, the rendered files will be written next to the source templates. If an output directory is provided, the source directory tree structure will be copied to the output directory with only the rendered files.

For instance, suppose we have the following directory structure :

.
└── test_dir
    ├── sub_dir_1
    │   ├── deep_dir
    │   │   └── template_1.txt.j2
    │   └── non_template_1.txt
    ├── sub_dir_2
    │   └── non_template_2.txt
    └── template_2.txt.j2

When we execute the command j2gpp ./test_dir/, we will get :

.
└── test_dir
    ├── sub_dir_1
    │   ├── deep_dir
    │   │   ├── template_1.txt
    │   │   └── template_1.txt.j2
    │   └── non_template_1.txt
    ├── sub_dir_2
    │   └── non_template_2.txt
    ├── template_2.txt
    └── template_2.txt.j2

But if we provide an output directory with the command j2gpp ./test_dir/ --outdir ./out_dir/, we will get :

.
├── test_dir
│   ├── sub_dir_1
│   │   ├── deep_dir
│   │   │   └── template_1.txt.j2
│   │   └── non_template_1.txt
│   ├── sub_dir_2
│   │   └── non_template_2.txt
│   └── template_2.txt.j2
└── out_dir
    ├── sub_dir_1
    │   └── deep_dir
    │       └── template_1.txt
    └── template_2.txt

We can also tell J2GPP to copy the non-template files with the command j2gpp ./test_dir/ --outdir ./out_dir/ --copy-non-template, then we will get :

.
├── test_dir
│   ├── sub_dir_1
│   │   ├── deep_dir
│   │   │   └── template_1.txt.j2
│   │   └── non_template_1.txt
│   ├── sub_dir_2
│   │   └── non_template_2.txt
│   └── template_2.txt.j2
└── out_dir
    ├── sub_dir_1
    │   ├── deep_dir
    │   │   └── template_1.txt
    │   └── non_template_1.txt
    ├── sub_dir_2
    │   └── non_template_2.txt
    └── template_2.txt

Or even to process non-templates files as templates anyway with the command j2gpp ./test_dir/ --outdir ./out_dir/ --render-non-template, then we will get :

.
├── test_dir
│   ├── sub_dir_1
│   │   ├── deep_dir
│   │   │   └── template_1.txt.j2
│   │   └── non_template_1.txt
│   ├── sub_dir_2
│   │   └── non_template_2.txt
│   └── template_2.txt.j2
└── out_dir
    ├── sub_dir_1
    │   ├── deep_dir
    │   │   └── template_1.txt
    │   └── non_template_1_j2gpp.txt
    ├── sub_dir_2
    │   └── non_template_2_j2gpp.txt
    └── template_2.txt

Supported formats for variables

Jinja2 supports variables types from python. The main types are None, Boolean, Integer, Float, String, Tuple, List and Dictionary. J2GPP provides many ways to set variables and not all types are supported by each format.

Command line define

Defines passed by the command line are interpreted by the Python ast.literal_eval() function which supports Python syntax and some additional types such as set().

j2gpp ./foo.c.j2 --define test_none=None             \
                          test_bool=True             \
                          test_int=42                \
                          test_float=3.141592        \
                          test_string1=lorem         \
                          test_string2="lorem ipsum" \
                          test_tuple="(1,2,3)"       \
                          test_list="[1,2,3]"        \
                          test_dict="{'key1': value1, 'key2': value2}" \
                          test_dict.key3=value3

YAML

test_none1:
test_none2: null

test_bool1: true
test_bool2: false

test_int: 42
test_float: 3.141592

test_string1: lorem ipsum
test_string2:
  single
  line
  text
test_string3: |
  multi
  line
  text

test_list1: [1,2,3]
test_list2:
  - 1
  - 2
  - 3

test_dict:
  key1: value1
  key2: value2
  key3: value3

JSON

{
  "test_none": null,

  "test_bool1": true,
  "test_bool2": false,

  "test_int": 42,
  "test_float": 3.141592,

  "test_string": "lorem ipsum",

  "test_list": [1,2,3],

  "test_dict": {
    "key1": "value1",
    "key2": "value2",
    "key3": "value3"
  }
}

XML

Note that XML expects a single root element. To avoid having to specify the root element when using the variables in a template, J2GPP automatically removes the root element level if it is named "_".

<_>
  <test_none></test_none>

  <test_bool1>true</test_bool1>
  <test_bool2>false</test_bool2>

  <test_int>42</test_int>
  <test_float>3.141592</test_float>

  <test_string>lorem ipsum</test_string>

  <test_list>1</test_list>
  <test_list>2</test_list>
  <test_list>3</test_list>

  <test_dict>
    <key1>value1</key1>
    <key2>value2</key2>
    <key3>value3</key3>
  </test_dict>
</_>

TOML

test_bool1 = true
test_bool2 = false

test_int = 42
test_float = 3.141592

test_string1 = "lorem ipsum"
test_string2 = """
multi
line
text"""

test_list = [1,2,3]

[test_dict]
key1 = "value1"
key2 = "value2"
key3 = "value3"

INI/CFG

Note that INI file expects data to be divided in sections with a header in square brackets. To avoid having to specify the root element when using the variables in a template, J2GPP automatically flattens the section whose header is "_".

[_]
test_bool1 = True
test_bool2 = False

test_int = 42
test_float = 3.141592

test_string = "lorem ipsum"

test_list = [1,2,3]

[test_dict]
key1 = value1
key2 = value2
key3 = value3

ENV

test_bool1 = True
test_bool2 = False

test_int = 42
test_float = 3.141592

test_string = lorem ipsum

test_list = [1,2,3]

test_dict = {'key1':'value1','key2':'value2','key3':'value3'}

CSV/TSV

CSV and TSV are interpreted as a list of objects with the same attributes. They are converted to a list of dictionaries whose name is the first cell of each line and the keys are the headers of each column.

CSV and TSV use the same loader, just with different delimiters. A different delimiter can be provided with the argument --csv-delimiter. To use the delimiter in a value, it can be escaped by defining an escape character with the argument --csv-escapechar, for instance the backslash "\".

By default, the whitespace around the keys and values in the CSV is stripped. This behaviour can be disabled with the argument --csv-dontstrip.

keys,key1,key2,key3
test_dict1,1,2,3
test_dict2,11,12,13
test_dict3,21,22,23

keys  key1  key2  key3
test_dict1  1  2  3
test_dict2  11  12  13
test_dict3  21  22  23

Scripting in J2GPP templates

An advanced use of the template feature and filters of Jinja2 and J2GPP allow this tool to do some amount of scripting. The main features allowing this usage are explained in this section.

Conditional and do extension

The basic Jinja2 feature of conditional blocks if/elif/else can be used alongside the do extension to more easily manipulate variables and complex objects such as Python dictionaries.

Note that, if possible, it is preferable to process the variables after loading and before rendering by providing a Python function with the --vars-post-processor command line argument.

Conditional and filters

As filters are Python functions, they can be very powerful, especially when coupled with the use of conditional blocks. However, Jinja2 optimizes processing by running some filters during the compilation phase, while conditional blocks are resolved at render time.

To fix this, you can use the @render_time_only decorator to force a filter or test to execute at render time only. This decorator is currently only available by installing a custom fork of Jinja2 :

git clone https://github.com/Louis-DR/jinja.git
cd jinja
pip3 install ./

Throw errors and warnings from template

If the version of Jinja2 installed supports the @render_time_only decorator, then the warning and error filters allow to throw warnings and erros from the template and display them in the J2GPP logs. This is useful with conditionals to perform data sanity checks for instance.

Writing files

The J2GPP filters write and append allow exporting the content of a block to another file. This can be used for a file combining elements contributed by multiple templates, for alternative versions of a file from a single template, for generating small annex files to a large template, for generating a files for each element in a list, etc. When coupled with the include or macro statements with nested temlates, it allows for even more complex outputs.

Note that if the installed Jinja2 version doesn't support the @render_time_only decorator, using the write and append filters in conditional blocks may results in unwnated behaviour.

Write example

For example, with the parent template parent.txt.j2:

This will be rendered to the parent output file
{% filter write("child.txt") %}
This will be rendered to the child output file
{% endfilter %}

>>> tree
.
└── parent.txt.j2

>>> j2gpp ./parent.txt.j2
[...]

>>> tree
.
├── child.txt
├── parent.txt
└── parent.txt.j2

>>> cat parent.txt
This will be rendered to the parent output file

>>> cat child.txt
This will be rendered to the child output file

Append example

If the file doesn't exists, the append filter will create it and be equivalent to write. However, if the file already exists, write will override it while append will append to the end of it. For example, with the parent template parent.txt.j2:

This will be rendered to the parent output file
{% filter append("child.txt") %}
This will be rendered to the child output file
{% endfilter %}

>>> tree
.
└── parent.txt.j2

>>> j2gpp ./parent.txt.j2
[...]

>>> tree
.
├── child.txt
├── parent.txt
└── parent.txt.j2

>>> cat parent.txt
This will be rendered to the parent output file

>>> cat child.txt
This will be rendered to the child output file

>>> j2gpp ./parent.txt.j2
[...]

>>> cat child.txt
This will be rendered to the child output file
This will be rendered to the child output file

>>> j2gpp ./parent.txt.j2
[...]

>>> cat child.txt
This will be rendered to the child output file
This will be rendered to the child output file
This will be rendered to the child output file

Path argument

The write and append filters require at least one argument, the name or path of the second file to generate. The path can be absolute, or relative to the generated file of the parent template. Non-existing directories will be created. For example, with the parent template parent.txt.j2:

{% filter write("foo/child.txt") %}
This will be rendered to the child output file
{% endfilter %}

>>> tree
.
└── src
    └── parent.txt.j2

>>> j2gpp ./src/parent.txt.j2 --outdir ./gen/
[...]

>>> tree
.
├── src
│   └── parent.txt.j2
└── gen
    ├── foo
    │   └── child.txt
    └── parent.txt

Writing to both files

By default, the content of the block is only written to the child file, and is not written to the parent rendered template. This behaviour can be changed by providing the filter argument preserve as True. For example, with the parent template parent.txt.j2:

{% filter write("child.txt", preserve=True) %}
This will be rendered to both parent and child output files
{% endfilter %}

>>> j2gpp ./parent.txt.j2 --outdir ./gen/
[...]

>>> cat ./parent.txt
This will be rendered to both parent and child output files

>>> cat ./child.txt
This will be rendered to both parent and child output files

Skipping the parent file

The source template can also be prevented from resulting in a generated file by providing the filter argument write_source as False, and only the content of write and append blocks will generate files. For example, with the parent template parent.txt.j2:

This will not be written to any file
{% filter write("child.txt") %}
This will be rendered to the child output file
{% endfilter %}

>>> tree
.
└── parent.txt.j2

>>> j2gpp ./parent.txt.j2
[...]

>>> tree
.
├── child.txt
└── parent.txt.j2

>>> cat child.txt
This will be rendered to the child output file