Internals Planning

Created 2017-04-29

  1. Zero config - cihai should be able to work without configuration with a default data backend.

  2. Incremental configuration cihai should be incrementally configurable, such as by specifying where data should be outputted.

  3. Relational backend cihai will use SQLAlchemy as a database backend to story information for retrieval.

  4. Automatic extension detection cihai will make data accessible to third party libraries if they exist in the script’s site-packages.

    e.g. If pandas is found, it will be able to return a pandas.DataFrame for a queried set of information.

  5. UNIHAN core cihai will use UNIHAN as a core and source of truth for information, as it contains all the glyphs and is reliable, free and well-maintained, and provides are good source of starter information.

  6. Data normalization cihai will adopt a standard data format to store additional CJK data sets within.

  7. Data liberation cihai libraries will be available under permissive licenses.

Zero config

cihai will be able to be used immediately without a user configuring their system.

cihai will conform with the XDG specification for determining where to check out data to. This includes:

  • Where to store downloaded source files, e.g. XDG_CACHE_HOME/cihai/downloads
  • Where store default backend data, e.g. XDG_DATA_DIRS/cihai/data, as well as the default file name used within, e.g. data.sqlite
  • Where to check for configuration files, e.g. XDG_CONFIG_HOME/cihai, as well as the default file name used within, e.g. cihai.yaml

These default directories will be where cihai will, by default, store information and search for configuration used in Incremental configuration.

Incremental configuration

The SQLAlchemy data backend used, which for SQLite, also includes the file path used to store the SQLite file, is customizable.

Relational backend

cihai will be bowered by a relational database backend.

Most python distributions include support for SQLite, which in conjunction with zero configuration, makes for data store that will work across a wide array of systems.

The data that cihai organizes will be primarily indexable by the glyph, and joined upon the glyph to pull in an ever expanding assortment of information on that character.

Automatic extension detection

Don’t reinvent the wheel, interoperate.

cihai will check for libraries such as pandas and other tabular libraries to easily produce native objects for the user based on their cihai data lookup.

This comes at no performance penalty since the ability to export in a third party object, such as a pandas.DataFrame.


cihai’s library of CJK information will be backed on the reliable UNIHAN database, which is a approved by the Unicode Consortium.


It is to be determined if UNIHAN will be vendorized in the packaging or retrieved remotely.

Data normalization

CJK datasets made available by cihai and contributors should follow, a yet to be determined, standard for keeping data conserved, readily available and sustainable.


The initial consideration, since 2013, was datasets would follow Data Packages.

In place of Data Packages, a simpler, and more lax guideline and alone with python interfaces may be considered. This determination is pending further review of datasets.

In place of frictionlessdata’s data package libraries, cihai may option for a simpler, yet more powerful system for making tabular data.

Data liberation

CC-0, MIT, ISC, BSD. Data sets should be available under licenses free from unintended side effects of derivitive creation.