Comments: precision about last cells.

%% Cell type:markdown id: tags:

# Production de tableaux pour correction des textes tokenisés sur Excel/Calc

Ce script permet de transformer des textes tokénisés en format XML-TEI avec des `//w[@lemma and @pos and @n]`.

Il nécessite des XML encodés selon le schéma ConDÉ, en version `base`.

Il comporte deux fonctions :

* get_w_txt(), qui construit les chaînes de caractères de chaque token,

* export_tokens_to_csv(), qui parse le XML, extrait les informations et ensuite les écrit dans un CSV.

Les dernières cellules donnent des emplacements pour lancer export_tokens_to_csv().

-----------------

# Production of tables to correct tokenized texts on Excel/Calc

This script transforms XML-TEI texts tokenized as `//w[@lemma and @pos and @n]`.

It needs XMLs encoded according to the ConDÉ `base` version schema.

It is written in two functions:

* get_w_txt(), which constructs strings out of each token,

* export_tokens_to_csv(), which parses the XML, extracts information and then writes it into a CSV file.

The last cells provide a place to launch export_tokens_to_csv().

%% Cell type:markdown id: tags:

#### Import libraries and declare TEI namespace

%% Cell type:code id: tags:

``` python

import xml.etree.ElementTree as ET

import csv

# Declaring base namespace (TEI),

# without a prefix (since it's the base namespace).

ET.register_namespace('', "http://www.tei-c.org/ns/1.0")

```

%% Cell type:markdown id: tags:

#### Extracting word strings from tokens

%% Cell type:code id: tags:

``` python

def get_w_txt(word):

    """

    Function constructing two strings from one word token (tei:w),

    first string being the 'diplomatic' version of the word,

    second string being the modernized/corrected version.

    If the word does not change, both strings will be identical.

    Function is based on the fact that each token has at least one child:

    it loops on all children of the word token.

    Function returns texte (list). texte[1] will always be the modernized/

    constructed version.

    If both constructed strings are identical, texte[0] will be an empty

    string. Otherwise it will contain the 'diplomatic' version.

    :param word: A parsed XML object corresponding to the following

    path: //tei:w.

    """

    # The future 'diplomatic' version.

    amorig = ''

    # The future 'modernized' version.

    expmod = ''

    # Storing TEI elements which may be treated the same.

    checklist = [

        '{http://www.tei-c.org/ns/1.0}height',

        '{http://www.tei-c.org/ns/1.0}supplied',

        '{http://www.tei-c.org/ns/1.0}c',

        '{http://www.tei-c.org/ns/1.0}hi'

    ]

    # If there is text before first child, add it to both versions.

    if word.text:

        amorig += word.text

        expmod += word.text

    # Looping on children of word token. According to the nature of

    # each child, treatment will be different.

    for item in word:

        # For all tei:height, tei:supplied, tei:c, tei:hi,

        # there is no modernization : add the text inside and

        # right after closing tag to both versions.

        if item.tag in checklist:

            amorig += item.text

            expmod += item.text

            if item.tail:

                amorig += item.tail

                expmod += item.tail

        # If child is a line beginning, there is no text inside:

        # if there is text directly afterwards, add it to both versions.

        elif item.tag == '{http://www.tei-c.org/ns/1.0}lb':

            if item.tail:

                amorig += item.tail

                expmod += item.tail

        # If child is a tei:choice element, then text will be different for

        # both versions. We parse its own children. Contents of tei:am and

        # tei:orig will go to 'diplomatic' version, while contents of

        # tei:expan and tei:reg will go to 'modernized' version. If there is

        # text directly after tei:choice element, add it to both.

        elif item.tag == '{http://www.tei-c.org/ns/1.0}choice':

            for subitem in item.findall('./*'):

                if subitem.tag == '{http://www.tei-c.org/ns/1.0}am' or subitem.tag == '{http://www.tei-c.org/ns/1.0}orig':

                    amorig += subitem.text

                elif subitem.tag == '{http://www.tei-c.org/ns/1.0}expan' or subitem.tag == '{http://www.tei-c.org/ns/1.0}reg':

                    try:

                        expmod += subitem.text

                    except:

                        "ALERTE !!"

            if item.tail:

                amorig += item.tail

                expmod += item.tail

        # If child is tei:add, then any of the previous children may be inside.

        # So we do the same there.

        elif item.tag == '{http://www.tei-c.org/ns/1.0}add':

            for subitem in item:

                if subitem.tag in checklist:

                    amorig += subitem.text

                    expmod += subitem.text

                    if item.tail:

                        amorig += subitem.tail

                        expmod += subitem.tail

                elif subitem.tag == '{http://www.tei-c.org/ns/1.0}lb':

                    if subitem.tail:

                        amorig += subitem.tail

                        expmod += subitem.tail

                elif subitem.tag == '{http://www.tei-c.org/ns/1.0}choice':

                    for subsub in subitem.findall('./*'):

                        if subsub.tag == '{http://www.tei-c.org/ns/1.0}am' or subsub.tag == '{http://www.tei-c.org/ns/1.0}orig':

                            amorig += subsub.text

                        elif subsub.tag == '{http://www.tei-c.org/ns/1.0}expan' or subsub.tag == '{http://www.tei-c.org/ns/1.0}reg':

                            expmod += subsub.text

                            if subitem.tail:

                                amorig += subitem.tail

                                expmod += subitem.tail

    # Construction of return list. If 'diplomatic' and 'modernized' versions

    # are different, then make a list with previous and latter. Otherwise,

    # replace previous with empty string.

    if amorig != expmod:

        texte = [amorig, expmod]

    else:

        texte = ['', expmod]

    return texte

```

%% Cell type:markdown id: tags:

#### Main function

%% Cell type:code id: tags:

``` python

def export_tokens_to_csv(chemin_entree, chemin_sortie):

    """

    Function parsing a tokenized TEI-XML file and returning a CSV

    table allowing manual correction outside of the XML file.

    Function uses the above get_w_txt() function to extract text

    from tokens containing children.

    :param chemin_entree: The local path to the TEI-XML file

        one needs to convert to a table.

    :param chemin_sortie: The local path to where one wants the

        resulting CSV table written.

    """

    # Counter to number table lines independantly from token numbers.

    l_count = 0

    # List used to store together //w/@n and //lb/@facs in document order,

    # so as to render the order of elements in the table.

    lbinitlist = []

    # Dictionnary used to store final information. Keys are w numbers //w/@n,

    # contents are all informations needed to make CSV file, including

    # //following::sibling:*[1][self::lb] or //child::lb when such is the case.

    winitdict = {}

    # Columns of final CSV.

    colonnes = [

        'n° de ligne',

        'nature',

        'n°/id',

        'forme "diplo"',

        'forme modernisée',

        'forme corrigée',

        'lemme',

        'pos',

        'à scinder',

        'à fusionner avec w n°',

        'à corriger',

        'corrigé'

    ]

    # Open and parse TEI XML.

    with open(chemin_entree) as infile:

        tree = ET.parse(infile)

        root = tree.getroot()

        # Loop on all elements containing //w children.

        # Add @n to lbinitlist if children are //w,

        # add @facs it children are //lb.

        # Order of elements in documents is preserved here.

        for parent in root.findall('.//*[{http://www.tei-c.org/ns/1.0}w]'):

            for child in parent.findall('./*'):

                if child.tag == '{http://www.tei-c.org/ns/1.0}w':

                    lbinitlist.append(child.get('n'))

                elif child.tag == '{http://www.tei-c.org/ns/1.0}lb':

                    lbinitlist.append(child.get('facs'))

        # Loop on //w elements. Register @n, @lemma and @pos and start

        # the dictionary entry in winitdict.

        for word in root.findall('.//{http://www.tei-c.org/ns/1.0}w'):

            texte = ''

            numero = str(word.get('n'))

            lemmes = str(word.get('lemma'))

            pos = str(word.get('pos'))

            winitdict[numero] = {'lemma':lemmes, 'pos':pos}

            # Then check if word has an //lb child (which would

            # not appear in lbinitlist) and, if so, add it to the dictionary

            # entry.

            if word.find('{http://www.tei-c.org/ns/1.0}lb'):

                winitdict[numero]['lb'] = word.find('{http://www.tei-c.org/ns/1.0}lb').get('facs')

            # Otherwise, check for the @n in lbinitlist, see if it is

            # followed by a //lb/@facs. If so, add it to the dictionary

            # entry. Otherwise, write 'None'.

            else:

                wIndex = lbinitlist.index(numero)

                try:

                    nextIndex = lbinitlist[wIndex+1]

                except:

                    nextIndex = ''

                if "_"in nextIndex:

                    winitdict[numero]['lb'] = nextIndex

                else:

                    winitdict[numero]['lb'] = 'None'

            # If word token has no children, we can get the

            # text directly. Otherwise, invoque get_w_txt() function

            # to compose it.

            if word.find('./*') == None :

                winitdict[numero]['original'] = ''

                winitdict[numero]['modernisé'] = word.text

            else:

                winitdict[numero]['original'] = get_w_txt(word)[0]

                winitdict[numero]['modernisé'] = get_w_txt(word)[1]

    # Open a CSV file in the output path, parse and write column headers.

    with open(chemin_sortie, 'w') as csv_file:

        csv_contenu = csv.DictWriter(csv_file, fieldnames = colonnes)

        csv_contenu.writeheader()

        # Then loop on word tokens in winitdict.

        for word in winitdict.keys():

            dicolocal = winitdict[word]

            l_count += 1

            # Check if token is ambiguous for sure and make the corresponding

            # variable.

            if '|' in dicolocal['pos'] or dicolocal['pos'] == 'Inconnu':

                a_corriger = 'X'

            else:

                a_corriger = ''

            # If current token dictionary contains a //lb/@facs, then

            # we add two lines: one for the line beginning, the other

            # for the actual word token. In-between, we add 1 to the

            # general counter.

            # Otherwise, we only add one line to the CSV table.

            if dicolocal['lb'] != 'None':

                csv_contenu.writerow({

                    'n° de ligne' : str(l_count),

                    'nature' : 'saut de ligne',

                    'n°/id' : dicolocal['lb'],

                    'forme "diplo"' :'',

                    'forme modernisée':'',

                    'forme corrigée':'',

                    'lemme' :'',

                    'pos' :'',

                    'à scinder' :'',

                    'à fusionner avec w n°' :'',

                    'à corriger' :'',

                    'corrigé' :''

                })

                l_count += 1

                csv_contenu.writerow({

                    'n° de ligne' : str(l_count),

                    'nature' : 'w',

                    'n°/id' : word,

                    'forme "diplo"' : dicolocal['original'],

                    'forme modernisée' : dicolocal['modernisé'],

                    'forme corrigée':'',

                    'lemme' : dicolocal['lemma'],

                    'pos' : dicolocal['pos'],

                    'à scinder' :'',

                    'à fusionner avec w n°' :'',

                    'à corriger' : a_corriger,

                    'corrigé' :''

                })

            else:

                csv_contenu.writerow({

                    'n° de ligne' : str(l_count),

                    'nature' : 'w',

                    'n°/id' : word,

                    'forme "diplo"' : dicolocal['original'],

                    'forme modernisée' : dicolocal['modernisé'],

                    'forme corrigée':'',

                    'lemme' : dicolocal['lemma'],

                    'pos' : dicolocal['pos'],

                    'à scinder' :'',

                    'à fusionner avec w n°' :'',

                    'à corriger' : a_corriger,

                    'corrigé' :''

                })

```

%% Cell type:markdown id: tags:

#### Where to launch the script for the whole collection

%% Cell type:code id: tags:

``` python

# If one wishes to treat several XML documents together,

# and they are named on the same model,

# and they are in the same directory,

# one can add each specific part of the filename

# into the following list.

temoins = ['bookname1', 'bookname2']

# Then one may add the actual paths and suffixes to this loop.

for temoin in temoins:

    export_tokens_to_csv('/path/to/folder/' + temoin + 'suffixe.xml',

                 temoin + '_tableau_pour_corrections.csv')

    # This allows the user to check which files are done yet.

    print(temoin + " : terminé")

```

%% Output

    basnage : terminé

%% Cell type:markdown id: tags:

#### Where to launch the script for one collection

%% Cell type:code id: tags:

``` python

export_tokens_to_csv(

    '/path/to/original.xml',

    '/path/to/table.csv'

)

```