Add a HumanErrorHandler

[funkyfuture]

this is a followup of https://github.com/nicolaiarocci/cerberus/issues/174, and here's a first design write-up:

The HumanErrorHandler produces text output that is read and interpreted by human beings. Most use-cases will propably inform users about invalid configurations and form data. Here are some ideas for features and options:

Value substitutions allow to replace document and schema constraint values. For example, a regex pattern that matches any URL would be replaced with "an URL".

The extent of communicated error description is configurable per error type. For example:

• Field 'foo' has an invalid data type. You provided an integer. Only strings are allowed.
• Field 'foo' has an invalid data type.
• Field 'foo' has an invalid data type. Only strings are allowed.

Similar errors over various fields are aggregated, eg:

• Fields 'foo' and 'bar' have invalid data types.

Some keywords can be replaced to match the application's context. For instance, the output may talk about a 'configuration file' instead of a 'document'.

Error messages can be localized.

The output might be aimed at terminal or html output and thus consider terminal / file line width or include tags and attributes.

In the end, this will not be trivial. But lesser demand wouldn't matter, given the targeted entities.

@luizpericolo @andreymal @Pithikos @k1-hedayati you might be interested in that.

addendum: a first implementation doesn't need to be feature-complete, but the anticipations must be reflected.

[andreymal]

Replacement of keywords instead of the exact phrase can lead to some problems with localization; for example, in Russian "configuration file" may be: "файл конфигурации", "файла конфигурации", "файлу конфигурации", "файлом конфигурации", "файле конфигурации"

Also "must be" may be:

• "должен быть" ("document must be" -> "документ должен быть");
• "должна быть" ("schema must be" -> "схема должна быть");
• "должно быть" ("field must be" -> "поле должно быть")
• or "должны быть" ("fields must be" -> "поля должны быть")

[funkyfuture]

what are these derivations caused by? are these nouns to be used in different cases (like lat: casus)?

differentiations of singular and plural is a must. would it be feasible to replace the noun and verbum spanning part? with agglutinating languages this idea propably won't work at all. maybe we should develop a localization framework first that also takes grammar as input. ;-)

please keep in mind that my native tongue is a cases-poor germanic one. in these regards it's a shame that the soviet union wasn't the avantgarde in computing. though it was propably easier to develop computing systems with a simple language as english.