pychronicles 0.1.5__tar.gz → 0.2.0__tar.gz

{pychronicles-0.1.5/pychronicles.egg-info → pychronicles-0.2.0}/PKG-INFO

@@ -1,7 +1,7 @@
-Metadata-Version: 2.1
+Metadata-Version: 2.4
 Name: pychronicles
-Version: 0.1.5
-Summary: A package for chronicle recognition
+Version: 0.2.0
+Summary: A package for chronicle recognition.
 Author-email: Thomas Guyet <thomas.guyet@inria.fr>
 Maintainer-email: Thomas Guyet <thomas.guyet@inria.fr>
 License:                    GNU LESSER GENERAL PUBLIC LICENSE
@@ -177,6 +177,11 @@ Keywords: data analysis,pattern,timed sequence,temporal data
 Classifier: License :: OSI Approved :: GNU Lesser General Public License v3 (LGPLv3)
 Classifier: Programming Language :: Python
 Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.7
+Classifier: Programming Language :: Python :: 3.8
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
 Classifier: Operating System :: OS Independent
 Requires-Python: >=3.7
 Description-Content-Type: text/markdown
@@ -197,6 +202,8 @@ Requires-Dist: build; extra == "build"
 Requires-Dist: twine; extra == "build"
 Provides-Extra: dev
 Requires-Dist: pytest; extra == "dev"
+Requires-Dist: pytest-cov; extra == "dev"
+Dynamic: license-file
 
 ---
 title: README
@@ -206,18 +213,19 @@ author: Thomas Guyet
 [![PyPI version](https://badge.fury.io/py/pychronicles.svg)](https://badge.fury.io/py/pychronicles)
 [![Downloads](https://static.pepy.tech/badge/pychronicles)](https://pepy.tech/project/pychronicles)
 [![GitHub license](https://img.shields.io/static/v1?label=Licence&message=LGPL-3&color=green)](https://www.gnu.org/licenses/lgpl-3.0.txt)
-
+[![coverage report](https://gitlab.inria.fr/tguyet/pychronicles/badges/master/coverage.svg)](https://gitlab.inria.fr/tguyet/pychronicles/-/commits/master)
+[![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/git/https%3A%2F%2Fgitlab.inria.fr%2Ftguyet%2Fpychronicles/HEAD?labpath=examples%2Fcarepathway_example.ipynb)
 
 # PyChronicles package
 
-A chronicle is a specification of the complex temporal behaviors as a graph of temporal constraints. More specifically, a chronicle is a multiset of events and a set of temporal constraints specifying that occurrences of pairs of events must occurs within a given temporal interval.
-It can be used to recognize complex behaviors in sequence the temporal events.
-PyChronicle is now compatible with Pandas and makes chronicle features [available on dataframes](#dataframe-accessor) representing timed sequences.
+A chronicle is a specification of complex temporal behaviors as a graph of temporal constraints. More specifically, a chronicle is a multiset of events and a set of temporal constraints specifying that the occurrences of pairs of events must occur within given temporal intervals.
+It can be used to recognize complex behaviors in sequences of temporal events.
+PyChronicles is now compatible with Pandas and makes chronicle features [available on dataframes](#dataframe-accessor) representing timed sequences.
 
-This library proposes a Python class to define a timed sequence, a chronicle, to read/save them in a standard CRS format. The main useful functionnality for chronicle is their efficient matching in a timed sequence, or pandas dataframes (including an accessor).
-The Abstracter class enables to create a chronicle from a collection of sequences.
+This library provides Python classes to define timed sequences and chronicles, and to read/save chronicles in the standard CRS format. The main functionality for chronicles is their efficient matching in timed sequences, or in pandas dataframes (through an accessor).
+The `Abstracter` class makes it possible to create a chronicle from a collection of sequences.
 
-For more details about the usage of this package, have a look at the [documentation](https://tguyet.gitlabpages.inria.fr/pychronicles/). And for more (formal) information about the temporal model of chronicle, see our [book](https://link.springer.com/book/10.1007/978-3-031-33693-5) and :
+For more details about the usage of this package, have a look at the [documentation](https://tguyet.gitlabpages.inria.fr/pychronicles/). For more (formal) information about the temporal model of chronicles, see our [book](https://link.springer.com/book/10.1007/978-3-031-33693-5):
 ~~~bibtex
 @book{chroniclesbook,
   title = {Chronicles: Formalization of a Temporal Model},
@@ -230,7 +238,7 @@ For more details about the usage of this package, have a look at the [documentat
 
 ## Package installation
 
-The [PyChronicles package](https://pypi.org/project/hydronaut) is available in [Python Package Index](https://pypi.org/) and can be installed using any standard Python package manager.
+The [PyChronicles package](https://pypi.org/project/pychronicles) is available in the [Python Package Index](https://pypi.org/) and can be installed using any standard Python package manager.
 ~~~sh
 pip install pychronicles
 ~~~
@@ -244,14 +252,14 @@ pip install -r requirements.txt
 pip install --upgrade .
 ~~~
 
-You can also use pip to install the repo from sources. For instance, run the following commands: 
+Alternatively, you can install the repository from sources with pip only: 
 ~~~sh
 git clone --recursive https://gitlab.inria.fr/tguyet/pychronicles
 cd pychronicles
 pip install .
 ~~~
 
-In case you want to visualize chronicles, you will need to install manually the `networkx` and `matplotlib`. By default, they are not installed and the plotting functions is disabled.
+In case you want to visualize chronicles, you need to install `networkx` and `matplotlib` manually. By default, they are not installed and the plotting function is disabled.
 ~~~sh
 pip install networkx matplotlib
 ~~~
@@ -259,20 +267,21 @@ pip install networkx matplotlib
 ## Basic Usage Example
 
 ```python
-import pychonicles
+import numpy as np
+from pychronicles import TimedSequence, Chronicle
 
 # create a timed sequence
 seq = [('a',1),('c',2),('b',3),('a',8)]
-ts = TimedSequence(np.array([e[2] for e in seq], dtype='float'), [e[0] for e in seq])
+ts = TimedSequence(np.array([e[1] for e in seq], dtype='float'), np.array([e[0] for e in seq]))
 
-#create a chronicle
+# create a chronicle
 c=Chronicle()
 c.add_event(0,'a')
 c.add_event(1,'b')
 c.add_constraint(0,1, (0.4,2.3))
 
 reco=c.match(ts)
-print(f"Does the chronicle matches the sequence? [{reco}]")
+print(f"Does the chronicle match the sequence? [{reco}]")
 
 reco=c.recognize(ts)
 print(f"What are the occurrences of the chronicle in the sequence? [{reco}]")
@@ -281,46 +290,46 @@ print(f"What are the occurrences of the chronicle in the sequence? [{reco}]")
 
 ## Efficient chronicle recognition
 
-The chronicle recognition package is a pure python package implemented using numpy features. To make it more efficient, it is possible to compile a cythonized version.
+The chronicle recognition package is a pure Python package implemented using numpy features.
 
 ### Timed Sequences
 
-A timed sequence is be created from different manners:
-* a pair of lists containing the events and the dates.
-* a simple list of items (`str`, `int` or `None`) with implicit timestamps : `['a', 'b', ..., None, 'c', None, 'd']`. In this case, `None` means that there is no event at the corresponding time instant.
-* a list of explicitly timestamped items (`str` or `int`) : `[ (1,'a'), (23,'b'), (30,'c'), (45, 'd')]`
+A timed sequence is created from a pair of arrays: a numpy array containing the dates, and an array containing the events (with the same length).
 
-The dates can be represented with dates (only with `np.datetime64` type) or floats. In case the dates are coded as `int`, they will be converted in `float`.
+The dates can be represented with dates (only with the `np.datetime64` type) or with floats. Integer dates must be converted into floats (the pandas accessor does it automatically).
 
-The type of events must be `str` or `int`.
+The type of the events must be `str` or `int`.
 
 _Usage Example_:
 ```python
 >>> seq = [('a',1),('c',2),('b',3),('a',8),('a',10),('b',12),('a',15),('c',17), ('b',20),('c',23),('c',25),('b',26),('c',28),('b',30)]
 >>> dates = np.array([np.datetime64('1970-01-01') + np.timedelta64(e[1],'D') for e in seq], dtype='datetime64')
 >>> data = np.array([e[0] for e in seq])
->>>  ts = TimedSequence(dates, data)
+>>> ts = TimedSequence(dates, data)
 ```
 
 ### Chronicles
 
-The core of the this package is the chronicle class that represents a chronicle temporal model which also offers efficient matching functionalities.
+The core of this package is the `Chronicle` class, which represents the chronicle temporal model and offers efficient matching functionalities.
 
 
-The chronicles handles two models of time (that must be consistent with the model of time in timed sequences):
-* discrete timestamps using floats
-* continuous timestamps using `datetime64` format. In this case the temporal constraints of a chronicle must be defined using `np.timedelta64` values.
+Chronicles handle two models of time (which must be consistent with the model of time of the timed sequences):
+* numerical timestamps using floats,
+* dates using the `datetime64` format. In this case, the temporal constraints of a chronicle must be defined using `np.timedelta64` values.
 
-A chronicle can not combine constraints of the two different kinds. The first defined constraint defines the model of time of the chronicle. 
+A chronicle cannot combine constraints of the two different kinds. The first defined constraint defines the model of time of the chronicle. 
 
 <br>
 
-The package implements efficient algorithms to recognize it. It benefits from numpy functionalities to increase their efficiency. There are three different ways to _recognize_ a chronicle in a sequence of a events:
+The package implements efficient algorithms to recognize chronicles. There are three different ways to _recognize_ a chronicle in a sequence of events:
+
+* the absence/presence recognition (`c.match(seq)`): its result is a boolean stating whether the chronicle occurs at least once in the sequence. This is the most efficient algorithm, as it stops as soon as an occurrence is found.
+* the occurrence enumeration (`c.recognize(seq)`): its result is the list of the occurrences of the chronicle in the sequence. Contrary to the first function, it looks for all the possible combinations of events. Thus, it is less efficient, but more informative.
+* the lazy occurrence enumeration (`c.iter_recognize(seq)`): it returns an iterator over the same occurrences, which are computed on demand. This is useful to process the occurrences one at a time, or to stop the enumeration early.
 
-* the absence/presence recognition (`c.match(seq)`): its result is a boolean stating whether the chronicle occur at least once in the sequence, this is the most efficient algorithm
-* the occurrence enumeration (`c.recognize(seq)`): its result is a list of occurrences of the chronicle in a sequence. Contrary to the first function, it looks for all possible combination of events. Thus it is less efficient, but more informative.
+In an occurrence, two identical events of a chronicle cannot match the same event of the sequence.
 
-Note that a chronicle is somehow similar to a simple temporal network and the set of constraints may be inconsistent or redundant. It is possible to _minimize_ the temporal constraints of a chronicle using the corresponding function.
+Note that a chronicle is somehow similar to a simple temporal network, and the set of constraints may be inconsistent or redundant. It is possible to _minimize_ the temporal constraints of a chronicle using the corresponding function.
 
 ```python
 >>> c=Chronicle()
@@ -353,16 +362,16 @@ You can also depict a chronicle as a matplotlib graph:
 
 ### Fuzzy chronicles
 
-The `FuzzyChronicle` class represents a class for approximated recognition of a chronicle. The chronicle is defined in the same way of a chronicle. 
-The temporal model is enriched with the modeling of a fuzzyness of temporal constraints (`lbda` parameter)
+The `FuzzyChronicle` class enables the approximate recognition of a chronicle. A fuzzy chronicle is defined in the same way as a chronicle.
+The temporal model is enriched with the modeling of the fuzziness of the temporal constraints (`lbda` parameter).
 
-In addition to the matching function presented above, it proposes a `cmp` function that finds occurrences of a chronicle with a degree of matching (parameter `threshold`).
+In addition to the matching functions presented above, it provides a `cmp` function that finds all the occurrences of a chronicle whose degree of matching is above a threshold (parameter `threshold`), together with their similarity. The `iter_cmp` function is its lazy version.
 
-More details about fuzzy chroniclesis available in the [article](https://hal.archives-ouvertes.fr/hal-03698361) (in French):
+More details about fuzzy chronicles are available in the [article](https://inria.hal.science/hal-02422796v1) (in French):
 ```bibtex
 @inproceedings{fchronicles,
   title = {{\'E}num{\'e}ration des occurrences d'une chronique},
-  author = {Guyet, Thomas and Besnard, Philippe and Ben Salha, Nasreddine and Samet, Ahmed and Lachiche, Nicolas},
+  author = {Guyet, Thomas and Besnard, Philippe and Samet, Ahmed and Ben Salha, Nasreddine and Lachiche, Nicolas},
   booktitle = {Actes de la conférence Extraction et Gestion des Connaissances},
   publisher = {{\'E}ditions RNTI},
   pages = {253--260},
@@ -372,29 +381,29 @@ More details about fuzzy chroniclesis available in the [article](https://hal.arc
 
 # Dataframe accessor
 
-The `pychronicles` package include an new accessor for pandas dataframe. 
-This accessor enables, denoted `tpattern` to use chronicle functionnalities directly with the pandas packages. In this case, you do not even need to manage your timed sequences by hand ... everything can be done directly with pandas.
+The `pychronicles` package includes a new accessor for pandas dataframes.
+This accessor, denoted `tpattern`, makes it possible to use chronicle functionalities directly with pandas. In this case, you do not even need to manage your timed sequences by hand: everything can be done directly with pandas.
 
 The pandas accessor enables:
 
 * to match or enumerate occurrences of a chronicle
-* to match an Metric Temporal Logic (MTL) formula
+* to match a Metric Temporal Logic (MTL) formula
 * to abstract a collection of sequences 
 * to extract the delays of matching occurrences of a chronicle
 
 
 ## Spirit of the accessor
 
-The spirit of our accessor is to use a pandas dataframe to encode a sequence or a collection of sequence. The index of the dataframe models the time ... it can be integer, float or date index. 
+The spirit of our accessor is to use a pandas dataframe to encode a sequence or a collection of sequences. The index of the dataframe models the time: it can be an integer, float or date index.
 
-A column of the dataframe acn be defined to specify the identifier of the sequence. In the following example, the column name `id` denotes this identifier. Intuitivelly, it can be used to make `groupby` operations, to process a collection of timed sequences. 
+A column of the dataframe can be defined to specify the identifier of the sequences. In the following example, the column named `id` denotes this identifier. Intuitively, it can be used to make `groupby` operations, to process a collection of timed sequences.
 
-The important feature of our package is that any other column can be added to the dataframe. The idea behind this modeling is that the dataframe gather all the information about the events. The columns describes each event (an can contain `None`). 
+The important feature of our package is that any other column can be added to the dataframe. The idea behind this modeling is that the dataframe gathers all the information about the events. The columns describe each event (and can contain `None`).
 
 
-The example below illustrates the representation of a collection of sequences with a Pandas dataframe. 
-The basic sequence is repeated three times to create a collection of three timed sequences. The index of the dataframe is defined with dates in this case. 
-The event names of the basic sequence feed the column `label`, and we added two columns:
+The example below illustrates the representation of a collection of sequences with a Pandas dataframe.
+The basic sequence is repeated three times to create a collection of three timed sequences. In this case, the index of the dataframe is defined with dates.
+The event names of the basic sequence feed the column `label`, and we added three columns:
 
 
 * a column `str_val` with the string values (twice the label)
@@ -402,7 +411,7 @@ The event names of the basic sequence feed the column `label`, and we added two
 * a column `id` that represents the identifier of a sequence.
 
 ```python
-#create a basic sequence
+# create a basic sequence
 seq = [('a',1),('c',2),('b',3),('a',8),('a',10),('b',12),('a',15),('c',17)]
 df = pd.DataFrame({
     "label": [e[0] for e in seq]*3,
@@ -416,19 +425,19 @@ df = pd.DataFrame({
 
 <br>
 
-Then, the definition of the event of a chronicle are no more the name of an event, but the description of a event of interest throught its description features. 
-When used with pandas dataframe, the events of a chronicle are queries on the dataframe.
+Then, the events of a chronicle are no longer event names, but descriptions of the events of interest through their features.
+When used with pandas dataframes, the events of a chronicle are queries on the dataframe.
 In the example above, an event could be defined by "`label=='a' & num_val>3`".
 
-At the time, our framework handles string and number attributes. The specification of an event uses the classical operators defined in the [Pandas documentation](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.query.html). 
+For the time being, our framework handles string and number attributes. The specification of an event uses the classical operators defined in the [Pandas documentation](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.query.html). 
 
-## Matching / enumerate occurrences of a chronicle
+## Matching / enumerating the occurrences of a chronicle
 
-Let us now use a dataframe defined as explained in the previous section with chronicle. 
+Let us now use a dataframe defined as explained in the previous section with a chronicle.
 
-For that, we first need to define a chronicle. It works exactly as before but, the definition of chronicle's events are queries.
+For that, we first need to define a chronicle. It works exactly as before, but the chronicle's events are defined by queries.
 
-For instance, the following example illustrates the definition a simple chronicle that uses queries as event:
+For instance, the following example illustrates the definition of a simple chronicle that uses queries as events:
 ```python
 c=Chronicle()
 c.add_event(0,'label=="a"')
@@ -438,24 +447,24 @@ c.add_constraint(0,1, (np.timedelta64(4,'D'),np.timedelta64(10,'D')))
 
 For more details about the syntax of queries, please refer to the [Pandas documentation](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.query.html). 
 
-This chronicle is semantically sound for being used in with the dataframe defined in the example of the previous section. 
+This chronicle is semantically sound to be used with the dataframe defined in the example of the previous section. 
 
 ### Dataframe representing a sequence
 
 In this section, we consider that the dataframe represents a unique sequence (for instance, we selected the first sequence of the previous dataframe with `df[df.id==1]`).
 
-The recognition functionalities are now accessible throught the Pandas dataframe:
+The recognition functionalities are now accessible through the Pandas dataframe:
 
-* `df.tpattern.match(c)` will return a boolean indicating whether the chronicle occurs in the sequence or not
-* `df.tpattern.recognize(c)` will return the list of occurrences of `c` in the sequence
-* `df.tpattern.project(c)` will return the list of delays for the temporal constraints of the chronicle `c`
+* `df.tpattern.match(c)` returns a boolean indicating whether the chronicle occurs in the sequence or not
+* `df.tpattern.recognize(c)` returns the list of the occurrences of `c` in the sequence (`df.tpattern.iter_recognize(c)` is its lazy version)
+* `df.tpattern.project(c)` returns a dataframe with the delays of the temporal constraints of the chronicle `c` in each of its occurrences
 
 Note that the `tpattern` keyword gives access to the features of our package.
 
 <br>
 
-Using chronicles becomes very easy with Pandas dataframe and it also gains in expressivity thanks to the use of queries as event. 
-We keep good computational efficiency of the recognition algorithms thanks to the use of the rewritting principle developed with semantic chronicles (see reference below). 
+Using chronicles becomes very easy with Pandas dataframes, and they also gain in expressivity thanks to the use of queries as events.
+We keep a good computational efficiency of the recognition algorithms thanks to the rewriting principle developed with semantic chronicles (see the reference below). 
 
 ```bibtex
 @inproceedings{semantic_chronicles,
@@ -469,20 +478,20 @@ We keep good computational efficiency of the recognition algorithms thanks to th
 
 ### Dataframe representing a collection of sequences
 
-To proceed with a dataframe containing several sequences, you can use the following idioms based on `apply`:
+To process a dataframe containing several sequences, you can use the following idiom based on `apply`:
 
 ```python
 reco=df.groupby('id').apply(lambda d: d.tpattern.match(c))
 ```
 
 In this case, the attribute `id` is used to select the events for each sequence, and then the `match` algorithm is applied to this selection.
-The result is a dataframe containing a boolean value for each sequence id.
+The result is a pandas series containing a boolean value for each sequence id.
 
-The same applies with `recognition` function.
+The same applies to the `recognize` function.
 
-## Use of MTL formula
+## Use of MTL formulas
 
-For situation recognitinon, there are strong connection between chronicles and metric temporal logics such as MTL or TPTL. See the [article](https://hal.archives-ouvertes.fr/hal-03777471):
+For situation recognition, there are strong connections between chronicles and metric temporal logics such as MTL or TPTL. See the [article](https://hal.archives-ouvertes.fr/hal-03777471):
 ~~~bibtex
 @inproceedings{chronicle_templogics,
   title = {Logical forms of chronicles},
@@ -493,12 +502,12 @@ For situation recognitinon, there are strong connection between chronicles and m
 }
 ~~~
 
-To illustrate that the principle of semantic chronicles can be broader, we extended it to the case of MTL formulae. This means that our framework enables to express a MTL formulae using query as events. 
+To illustrate that the principle of semantic chronicles can be broader, we extended it to the case of MTL formulas. This means that our framework makes it possible to express MTL formulas using queries as events.
 
 
-The `tpattern` accessor provides a function `match_mtl` that is dedicated to the recognition of a MTL formulae in the timed sequence. This function returns a boolean value. There is not equivalent to the enumeration of occurrences of a chronicle.
+The `tpattern` accessor provides a function `match_mtl` that is dedicated to the recognition of an MTL formula in the timed sequence (only for dataframes indexed by floats or integers). This function returns a boolean value. There is no equivalent to the enumeration of the occurrences of a chronicle.
 
-The syntax of an MTL formula is based on the two operators `F` (eventually) and `G` (globally). These operators can be specified with temporal constraints (in brackets). They combines events defined in the same way as for chronicle events.
+The syntax of an MTL formula is based on the two operators `F` (eventually) and `G` (globally). These operators can be specified with temporal constraints (in brackets). They combine events defined in the same way as chronicle events.
 
 ```python
 query = ' F(label=="a" & F[2.9,5]( label=="b" & num_val>5 ))'
@@ -509,13 +518,13 @@ This feature is based on the [Python implementation of MTL](https://github.com/m
 
 ## Abstraction of a sequence
 
-Finally, we also provide a function that enables to abstract a collection of sequences into a chronicle. 
+Finally, we also provide a function that abstracts a collection of sequences into a chronicle.
 
-Contrary to the recognition functions that can handle several columns, the abstraction requires to define a column that correspond to the event name. In addition, it requires to define the attribute name that correspond to the identifier of the sequences. 
-Then, the function, named `abstract` takes two arguments and outputs a single chronicle that occurs in all the sequences.
+Contrary to the recognition functions that can handle several columns, the abstraction requires a column that corresponds to the event names. In addition, it requires the name of the attribute that corresponds to the identifier of the sequences.
+Then, the function, named `abstract`, takes these two arguments and outputs a single chronicle that occurs in all the sequences.
 
 
-The following example extracts a chronicle from the dataframe introduced above. It focuses event on the label attribute.
+The following example extracts a chronicle from the dataframe introduced above. Its events are based on the `label` attribute.
 ```python
 chro = df.tpattern.abstract('label', 'id')
 ```
@@ -524,7 +533,7 @@ chro = df.tpattern.abstract('label', 'id')
 
 ## Export / import of chronicles
 
-It is possible to specify chronicles using the CRS format. The following code illustrate the syntax for specifying a chronicle in this format.
+It is possible to specify chronicles using the CRS format (see the functions `load` and `to_crs` of the `pychronicles.utils` module). The following code illustrates the syntax for specifying a chronicle in this format.
 
 ```
 chronicle C27_sub_0[]()
@@ -545,16 +554,16 @@ chronicle C27_sub_0[]()
 
 * implementation of chronicle mining algorithms
 * graphical chronicle editor
-* better cythonize the recognition
+* compile the recognition (e.g. with Cython)
 * optimized matching of several chronicles at the same time
 
 
-If you think that this package can be improved ... do not hesitate to contact the author to request additional features. He will do his best to maintain the code and improve it continuously.
+If you think that this package can be improved, do not hesitate to contact the author to request additional features. He will do his best to maintain the code and improve it continuously.
 
 
 # Authorship
 
 * **Author:** Thomas Guyet
 * **Institution:** Inria/AIstroSight
-* **date:** 08/2024
+* **date:** 10/2026