sderose / pythonlibs Goto Github PK

#README file for PYTHONLIBS repo#

This repository includes a bunch of Python libraries used by my other repos, or generally useful (or at least, I think so).

It is gradually narrowing in scope, to mainly be pluggable subclasses that (imho) improve on standard Python libraries.

##Highlights##

• BlockFormatter.py: A very basic Markdown-aware formatter to use with Python argparse. This version basically preserves breaks and blank lines for paragraphs and lists. A forthcoming version will do inlines too.

Many items in this repo also have Perl versions in [https://github.com/sderose/PERLLIBS].

Several items here have moved or will be moving to [https://github.com/sderose/XML] or other neighboring repos. See the end of this README for more information.

• fsplit.py: Like Python's csv class, but many more features and capabilities.

• PowerWalk.py -- A considerably more flexible version of os.walk.

• argparsePP.py: A set of extensions to Python's argparse, supporting several additional datatypes (probability; XSD types; number in user's choice of octal, hex, or decimal; regex restrictions, etc. Also provides easy creation of both positive and negative sides for Boolean options, or promoting the choices of enum options to their own options; and case-ignoring.

• CharDisplay.py: Show a lot of information about any Unicode code point.

##Short descriptions##

("meh" below indicates the item is unfinished or imho uninteresting)

• BlockFormatter.py -- a drop-in replacement for the argparse formatters. Python's default one wraps everything into one massive blob of text; and its "Raw" one does no wrapping at all. BlockFormatter handles rudimentary but very common conventions:

** blank lines are retained (say, between paragraphs) ** indented lines, or lines starting with , #, =, etc. keep the newline and whitespace before them. ** unindented lines get wrapped together with the preceding line, and the whole block is indented to match the first line's indent (if any). ** by default, a 2-column hanging indent is adde to non-first lines of indented blocks (beyond the indentation expressed by their first line). This is so that list items flagged by "" etc have clearer boundaries. ** The whole thing is wrapped to the current window width.

Try it, it makes argparse help much better. Also see BlockFromatterPlus, which adds support for inline markup and in-terminal bold, italic, and other capabilities; as well as MarkupHelpFormatter, which is still in development but has even more flexible formatting.

• CharDisplay.py -- given a character code as 999, 0xFFFF, 0177, etc., show a ton of information about that Unicode code point. [Moving]

• ColorManager.py -- provides ANSI color control, interfaced via simple names like red/white/bold (for bold red text on a white background). [Moving]

• DictTuple.py -- A subclass of dict that provides a number of features from defaultdict and namedtuple as well. It can even enforce types for members, allow or prohibit None for members, etc.

• DomExtensions.py -- Everything you always wanted to have in DOM, but were afraid to ask for. Has a method to monkey-patch it all onto a DOM implementation. [Moving]

• Homogeneous.py -- Defines 'hlist' and 'hdict', subclasses of list and dict that let you constrain values (and/or keys for hdict) to certain types and/or values. Whether subclasses are ok, and whether None is ok, also controllable.

• LooseDict.py -- A subclass of dict, that lets you attach a "normalization" function for keys. For example, specifying lower() gives you a case-ignoring dictionary. Each entry stores both the normalized and (last-set) unnormalized key.

• MarkupHelpFormatter.py -- [this is in process, and may be broken at various times]. A text-formatter that can be attached to Python argparse, so you can use MarkDown, MediaWiki, POD, etc. to get formatted help. Makes heavy use of ANSI color and Unicode variations for pretty formatting (you want bold italic sans serif, or bold Fraktur, on your terminal? You got it. See also [https:github.com/sderose/Charsets/ord] --math).

• PowerWalk.py -- A more flexible version of os.walk. This provides much (not all) the functionality of 'grep' and 'find' for choosing files; can open tar and gzip files directly, and provides even easier iteration than os.walk.

• ReadAny.py -- meh

• SimplifyUnicode.py -- meh

• TabularFormats.py -- [kinda broken right now -- check out fsplit.py]. supports several "tabular" data formats, including many variants of CSV. This started out as a port from Perl, but needs to be updated so it sits on top of exiting Python libs where practical. See also fsplit.py.

• Tokenizer.py -- a very Unicode-aware word tokenizer for NLP work.

• TokensEN.py -- includable by Tokenizer.py to support English-specific stuff like contractions.

• alogging.py -- a logging package, mostly upwardly compatible with the usual successive indentation, etc.

• argParsePP.py -- half-done upgrade to argparse, for easier support of -x vs. --x, --no-x, and other conventions.

• fsplit.py -- a variation on Python's string split() method and the fairly similar csv library. The API is almost an exact superset of csv. Perhaps I should rename this 'csvPP' or something?

In brief, fsplit supports all the options of Python's csv package and of string split() itself: (escaping, quoting, doubling, lstrip, header records, etc), as well as these options and more:

** Multi-character delimiters (split() but not csv offers those) -- no more re-mapping before you can read ** More than one allowed quotechar (for example, single and double) -- no more writing r"('[^']'|"[^\"]")" all the time ** Unicode quote characters (curly, angle, etc) -- no more "uneducating" quotes in a CSV you edited ** Special character escapes like \xFF, \uFFFF, �, •, etc. -- no more decoding before you can read these ** Option to ignore repeated delimiters -- an alternate you sometime see, that few thing handle. ** Type-checking -- catch more data errors by saying what you expect -- pass your own types as constraints for particular fields ** Automatic type detection and casting ("3.14" returned as a float, etc.) -- no more casting everything on input -- prettier output -- complex numbers, too ** Formatting output files -- CSV in columns you can read ** Definable comment delimiter -- CSV files you can document

• sjdUtils.py -- very generally-applicable stuff, like indenting and escaping and unescaping XML, JSON; making Unicode and control characters printable, etc.

##Files moved or to be moved##

###To Charsets###

• CharDisplay.py: Get lots of information on Unicode characters

###To Color###

• ColorManager.py: Easy interface to ANSI terminal colors and effects.

###To NLP###

• Tokenizer.py: A pretty good regex-based English NLP tokenizer.

###To XML###

• XmlOutput.py: A package to make writing WF XML easy. Methods like openElement('p', { 'id':"foo", "class":2 }), openElement([ 'html', 'body', 'div', 'p']), closeToElement('body'), queueAttribute(), along with automatic and manual escaping, indentation options, can be pretty handy. sjdUtils has commonly-needed small methods for escaping, unescaping, and pretty-printing XML, JSON, Unicode, regexes, etc.; hooks for colorizing and logging via other packages; padding strings and aligning arrays into columns; formatting numbers with commas or as K/M/G/T/P millenia suffixes rather than lots of digits; formatting dates and times; etc.

• DomExtensions.py: Everything you always wanted to call in DOM, but didn't implement. In particular, it adds a lot of methods modelled on XPath (are you surprised?).

• ElementManager.py -- Manage an XML schema (very basic) [Moving]

• EntityManager.py -- Manage XML data sources, catalogs, etc. [Moving]

• XmlRegexes.py (now in XML/PARSERS/XmlConstructs.py).

• XmlNlpLoad.py

• XmlOutput.py -- helps write well-formed XML with simple calls like openElement(name, attrs), closeThroughElement(), escaping, queueAttribute(), and so forth. Maintains a stack so you know if you break things.

• XmlTuples.py -- Supports a very simple, XML-conforming 3-element schema for CSV-like data. Easily interconvertible with most any other such format.