searchanddisplace-ingest/app/Parser/DocxParser/ParseDocx.php

<?php

namespace App\Parser\DocxParser;

use App\Parser\DocxParser\Traits\Helper;
use Illuminate\Support\Facades\Log;
use PhpOffice\PhpWord\IOFactory;
use function GuzzleHttp\Psr7\str;

class ParseDocx
{

    use Helper;

    protected $currentNumberingIndex = 1;


    public function fromUploadedFile($file)
    {
        try {
            $docxFileLoader = IOFactory::load($file);
            Log::info('Parse docx');

            return $this->parseLoadedDocx($docxFileLoader);
        } catch (\Exception $exception) {
            dd($exception);
            throw new \Exception($exception->getMessage());
        }

    }


    private function parseLoadedDocx($docx)
    {
        $styles = 0;
        foreach ($docx->getSections() as $page) {

            $handler    = $this->getHandler($page);
            $paragraphs = $handler->handle($page);
            if ($paragraphs) {
                foreach ($paragraphs as $index => $paragraph) {
                    try {
                        if ($paragraph && $paragraph[ 'type' ] !== 'textBreak' && (isset($paragraph[ 'content' ][ 'type' ]) && $paragraph[ 'content' ][ 'type' ] !== 'textBreak') || $paragraph[ 'type' ] == 'table') {
                            $result[] = $paragraph;
                            if (isset($paragraph[ 'styleName' ])) {
                                $styles++;
                            }
                        }
                    } catch (\Exception $e) {
                        dd($e);
                    }

                }
            }
        }
        $depthTypeType = count($result) / 2 <= $styles ? 'styleDepth' : 'depth';

        return $this->setTheNumbering($result, null, $depthTypeType);
    }


    private function setTheNumbering($paragraphs, $parentNumbering = null, $depthType = 'depth')
    {
        $result     = [];
        $paragraphs = $this->buildTheChildrens($paragraphs, $depthType);
        for ($index = 0; $index < count($paragraphs); $index++) {
            $paragraph = $paragraphs[ $index ];
            try {
                if ($paragraph[ 'type' ] !== 'table' && ($paragraph[ $depthType ] === 0 || $parentNumbering) && strpos($paragraph[ 'styleName' ],
                        'BodyText') === false) {

                    $paragraph[ 'content' ][ 'numbering' ]     = ($parentNumbering) ? $parentNumbering.((int) $index + 1).'.' : $this->currentNumberingIndex.'.';
                    $paragraph[ 'content' ][ 'numbering_row' ] = ($parentNumbering) ? ((int) $index + 1) : $this->currentNumberingIndex;

                    if ($paragraph[ 'children' ] && count($paragraph[ 'children' ])) {
                        $paragraph[ 'children' ] = $this->setTheNumbering($paragraph[ 'children' ],
                            $paragraph[ 'content' ][ 'numbering' ], $depthType);

                    }

                    if (! $parentNumbering) {

                        $this->currentNumberingIndex++;
                    }


                } elseif (isset($paragraph[ 'content' ][ 'numbering' ]) && isset($paragraph[ 'children' ]) && count($paragraph[ 'children' ])) {
                    $paragraphs[ $index ] = $this->setChildrenNumbering($paragraphs[ $index ]);
                } elseif (isset($paragraphs[ $index ][ 'content' ][ 'numbering' ]) && isset(last($result)[ 'content' ][ 'numbering' ]) && $paragraphs[ $index ][ 'content' ][ 'numbering' ] == last($result)[ 'content' ][ 'numbering' ]) {


                }
            } catch (\Exception $e) {
                dd($e);
            }
            $result[] = $paragraphs[ $index ];

        }

        return $result;
    }


    /**
     * @param $parent
     *
     * @return mixed
     */
    private function setChildrenNumbering($parent)
    {

        $numbering = 1;
        for ($j = 0; $j < count($parent[ 'children' ]); $j++) {
            $children = $parent[ 'children' ][ $j ];

            if ($children[ 'type' ] == 'listItemRun' || isset($children[ 'content' ][ 'numbering' ])) {
                $parentNumber                                           = $parent[ 'content' ][ 'numbering' ];
                $parent[ 'children' ][ $j ][ 'content' ][ 'numbering' ] = (substr(trim($parentNumber),
                        strlen(trim($parentNumber)) - 1) == '.') ? $parentNumber.$numbering : $parentNumber.'.'.$numbering;
                if (count($parent[ 'children' ][ $j ][ 'children' ])) {

                    $parent[ 'children' ][ $j ] = $this->setChildrenNumbering($parent[ 'children' ][ $j ]);
                }

                $numbering++;
            }
        }

        return $parent;
    }


    /**
     * @param $paragraphs
     *
     * @return array
     */
    private function buildTheChildrens($paragraphs, $depthType)
    {
        $alreadyHandledIndexes = [];
        $result                = [];

        for ($i = 0; $i < count($paragraphs); $i++) {

            if (in_array($i, $alreadyHandledIndexes)) {
                continue;
            }
            $j = $i + 1;

            for ($j; $j < count($paragraphs); $j++) {

                if (in_array($j, $alreadyHandledIndexes)) {
                    continue;
                }

                if (isset($paragraphs[ $j ][ 'content' ][ 'content' ]) && $paragraphs[ $j ][ 'content' ][ 'content' ] === '<p></p>') {
                    $alreadyHandledIndexes[] = $j;
                    $j++;
                }

                if (isset($paragraphs[ $i ][ $depthType ]) && isset($paragraphs[ $j ][ $depthType ]) && $paragraphs[ $i ][ $depthType ] !== null && $paragraphs[ $j ][ $depthType ] !== null && $paragraphs[ $i ][ $depthType ] < $paragraphs[ $j ][ $depthType ]) {

                    $paragraphs[ $i ] = $this->handlePossibleChild($paragraphs[ $i ], $paragraphs[ $j ], $i,
                        $depthType);


                } elseif (isset($paragraphs[ $j ][ 'styleName' ]) && $paragraphs[ $j ][ 'styleName' ] === 'ListParagraph' && $paragraphs[ $i ][ $depthType ] === null && substr(strip_tags($paragraphs[ $i ][ 'content' ][ 'content' ]),
                        -1) === ':') {
                    $paragraphs[ $i ] = $this->handlePossibleChild($paragraphs[ $i ], $paragraphs[ $j ], $i,
                        $depthType);

                } elseif (isset($paragraphs[ $j + 1 ]) && isset($paragraphs[ $j + 1 ][ 'content' ][ 'content' ]) && isset($paragraphs[ $j ]) && isset($paragraphs[ $j ][ 'content' ][ 'content' ]) && substr(strip_tags($paragraphs[ $j ][ 'content' ][ 'content' ]),
                        -1) === ':' && (isset($paragraphs[ $j + 1 ]) && ctype_lower(substr(trim(strip_tags($paragraphs[ $j + 1 ][ 'content' ][ 'content' ])),
                            0,
                            1)) || (isset($paragraphs[ $j + 1 ]) && substr(trim(strip_tags($paragraphs[ $j + 1 ][ 'content' ][ 'content' ])),
                                strlen(trim(strip_tags($paragraphs[ $j + 1 ][ 'content' ][ 'content' ]))) - 1) == ';'))) {
                    $k                       = $j + 1;
                    $alreadyHandledIndexes[] = $k;
                    while (isset($paragraphs[ $k ]) && substr(str_replace('and', '',
                            trim(strip_tags(str_replace('and', '', $paragraphs[ $k ][ 'content' ][ 'content' ])))),
                            strlen(str_replace('and', '', trim(strip_tags(str_replace('and', '',
                                $paragraphs[ $k ][ 'content' ][ 'content' ]))))) - 1) == ';') {
                        $paragraphs[ $j ][ 'children' ][] = $paragraphs[ $k ];
                        $alreadyHandledIndexes[]          = $k++;

                    }

                    $paragraphs[ $i ] = $this->handlePossibleChild($paragraphs[ $i ], $paragraphs[ $j ], $i,
                        $depthType);


                } elseif (isset($paragraphs[ $i ][ 'styleName' ]) && $paragraphs[ $i ][ $depthType ] !== $paragraphs[ $j ][ $depthType ] && strpos($paragraphs[ $i ][ 'styleName' ],
                        'Heading2') !== false && ((isset($paragraphs[ $j ][ 'depth' ]) || ($paragraphs[ $j ][ 'type' ] == 'textRun' && isset($paragraphs[ $j ][ 'content' ][ 'numbering' ])) && is_null($paragraphs[ $j ][ 'styleName' ])))) {

                    $paragraphs[ $i ] = $this->handlePossibleChild($paragraphs[ $i ], $paragraphs[ $j ], $i,
                        $depthType);


                } else {

                    break;
                }

                $alreadyHandledIndexes[] = $j;

            }
            $result[]                = $paragraphs[ $i ];
            $alreadyHandledIndexes[] = $i;

        }

        return $result;
    }


    /**
     * @param $parent
     * @param $child
     * @param $i
     *
     * @return mixed
     */
    private function handlePossibleChild($parent, $child, $i, $depthType)
    {

        // Must iterate through parent children
        if (isset($parent[ 'children' ]) && count($parent[ 'children' ]) === 0) {
            if ($parent[ $depthType ] < $child[ $depthType ] || $parent[ $depthType ] === null) {
                $parent[ 'children' ][] = $child;
            } elseif (strpos($parent[ 'styleName' ],
                    'Heading') !== false && isset($child[ 'content' ][ 'numbering' ]) && substr_count($child[ 'content' ][ 'numbering' ],
                    '.') == 1) {
                $parent[ 'children' ][] = $child;
            } else {
                return $parent;
            }

            return $parent;
        }

        $lastParentChild = last($parent[ 'children' ]);
        // Possible to be either child or grandchild
        if ($lastParentChild[ $depthType ] && $child[ $depthType ] > $lastParentChild[ $depthType ]) {

            $lastParentChild = $this->handlePossibleChild($lastParentChild, $child, $i, $depthType);

        } else {

            if ($child[ $depthType ] === $lastParentChild[ $depthType ]) {
                $parent[ 'children' ][] = $child;

                return $parent;
            }

            if (((isset($lastParentChild[ 'styleDepth' ]) && $lastParentChild[ 'styleDepth' ] === $child[ 'depth' ])) && $lastParentChild[ 'index' ] !== $child[ 'index' ]) {

                $parent[ 'children' ][] = $child;

                return $parent;
            }
        }

        $parent[ 'children' ][ count($parent[ 'children' ]) - 1 ] = $lastParentChild;

        return $parent;

    }

}