Exceptions and logging¶

Exceptions¶

All library exceptions inherit the base class EstsError and a built-in Python class, so they can be caught by either, and except ValueError keeps working.

Exception	Built-in class	When raised
`EstsError`	`Exception`	Base class, never raised itself
`SourceTypeError`	`TypeError`	The source is not a string or a `Doc`; the text of an extractor is not a string; a string, a `Doc`, an iterator or an object that is not iterable is passed where a list of words is expected, or the list holds something other than strings, such as spaCy tokens; the texts of a visualizer are not lists of words; the frequencies of `zipf` are not a `Counter`; the measure of `fingerprinting` or the tokenizer is not callable, or the tokenizer returns a non-iterable or something other than strings; a path is not a string or a `Path`; the stopwords or the clichés (style metrics, highlighting) are a string, not a list; a word of the syllables, of the transcription or of the stopwords is not a string; the keywords of `keyness_plot`, the collocations of `collocation_network` or the concordances of `format_kwic` are not a list; the features of `compare_corpora` or `corpus_features` are not a function
`SourceError`	`ValueError`	The source has no words, sentences, texts or collocations, lacks the annotation a statistic needs (parts of speech, lemmas, dependency parse), is a string longer than the `max_length` of the pipeline (for the verbal nouns, has a sentence longer than it), or nothing is left after culling; Delta or the principal components get fewer than three texts; a matrix of distances is not square or has an infinite distance; the pipeline of `function_words_profile` does not tag parts of speech; no text of a comparison has a window of enough words; the keyword of `wordtree` is not found or has no word next to it
`ParameterError`	`ValueError`	Not an integer: a length, window, step, number of items or other count, even a whole float. Out of range: a threshold, window, segment size, number of items, logarithm base, confidence level, number of bootstrap samples or bound of a frequency band (1-10,000); a negative limit of records or length bound of an extractor; parts of `dispersion` that do not add up to the words; an empty keyword of `kwic`. Unknown: a preset, scale, metric name, measure, variant, keyword field, genre or part of speech. Highlighting: an unknown layer or one the source does not allow (syntactic layers and verbal nouns on a string or on a `Doc` without the parse or the lemmas), layers that are neither a list nor a string, an alliteration threshold outside (0, 1], a number of words of a long sentence or of syllables of a complex word below 1. Plots: the axes are not a matplotlib `Axes`. A reading ease that is not a finite number (`nan`, an infinity) converted into a grade; parts of speech of `find_connectors` that are not as many as the words
`UnknownStatError`	`ParameterError`, `KeyError`	An unknown statistic is requested by name, as in `DiversityStats.windowed`
`DatasetNotFoundError`	`OSError`	The spaCy model is not installed or a dataset is not downloaded; the message gives the command that fixes it
`DataFileError`	`ValueError`	The archive of a dataset is not ZIP or TAR, cannot be extracted, is empty or has paths outside its directory, or its target directory cannot be created; a line of a dataset file cannot be read
`DownloadError`	`RuntimeError`	The file could not be downloaded or its directory created, or it failed the checksum twice

The classes are available from ests and from ests.exceptions. They are the classes of the anyTS core under the same names, EstsError being its anyts.exceptions.AnyTSError: aliases, not subclasses, so except EstsError catches the errors raised by the code of the core too, and a traceback shows anyts.exceptions.SourceTypeError.

Note

DatasetNotFoundError is the first error a new user usually meets: MorphStats("El gato duerme") without the model es_core_news_sm raises it with the download command. SpanishLiterature().get_texts() before download() and the statistics of LexicalStats by the frequency dictionary raise it too.

CodeResult

from ests import EstsError, ParameterError, WordsExtractor

try:
    WordsExtractor(ngram_range=(2, 1))
except ParameterError as e:
    print(e)

try:
    WordsExtractor(tokenizer=42).extract("El gato duerme.")
except EstsError as e:
    print(type(e).__name__)

The lower N-gram bound is greater than the upper
SourceTypeError

Logging¶

The library prints nothing on its own: the anyTS core sends messages about downloading and extracting datasets (the download() methods) to the anyts logger at the INFO level. The logger has a NullHandler by default, so the messages are silent; to see them, configure logging in your application:

CodeResult

import logging

from ests.datasets import SpanishSonnets

logging.basicConfig(level=logging.INFO, format="%(name)s: %(message)s")
SpanishSonnets(data_dir="/tmp/ests_data").download()

anyts.datasets: Downloading the file https://github.com/SergeyShk/esTS/raw/master/ests/datasets/data/spanish_sonnets_v1.tar.xz
anyts.datasets: The file is downloaded: /tmp/ests_data/spanish_sonnets_v1.tar.xz
anyts.datasets: Extracting the archive /tmp/ests_data/spanish_sonnets_v1.tar.xz

The print_stats() methods of the statistics classes, MorphStats.print_markers() and print_kwic() write to standard output by design; logging does not affect them.