diff options
Diffstat (limited to 'src/scraping/buxton/scraper.py')
| -rw-r--r-- | src/scraping/buxton/scraper.py | 43 |
1 files changed, 33 insertions, 10 deletions
diff --git a/src/scraping/buxton/scraper.py b/src/scraping/buxton/scraper.py index 700269727..182b22a1a 100644 --- a/src/scraping/buxton/scraper.py +++ b/src/scraping/buxton/scraper.py @@ -1,4 +1,5 @@ import os +from shutil import copyfile import docx2txt from docx import Document from docx.opc.constants import RELATIONSHIP_TYPE as RT @@ -15,7 +16,9 @@ source = "./source" dist = "../../server/public/files" db = MongoClient("localhost", 27017)["Dash"] +target_collection = db.newDocuments schema_guids = [] +common_proto_id = "" def extract_links(fileName): @@ -84,7 +87,7 @@ def write_schema(parse_results, display_fields, storage_key): "height": 600, "panX": 0, "panY": 0, - "zoomBasis": 0.5, + "zoomBasis": 1, "zIndex": 2, "libraryBrush": False, "viewType": 2 @@ -92,7 +95,7 @@ def write_schema(parse_results, display_fields, storage_key): "__type": "Doc" } - fields["proto"] = protofy("collectionProto") + fields["proto"] = protofy(common_proto_id) fields[storage_key] = listify(proxify_guids(view_guids)) fields["schemaColumns"] = listify(display_fields) fields["backgroundColor"] = "white" @@ -106,8 +109,8 @@ def write_schema(parse_results, display_fields, storage_key): fields["isPrototype"] = True fields["page"] = -1 - db.newDocuments.insert_one(data_doc) - db.newDocuments.insert_one(view_doc) + target_collection.insert_one(data_doc) + target_collection.insert_one(view_doc) data_doc_guid = data_doc["_id"] print(f"inserted view document ({view_doc_guid})") @@ -136,7 +139,7 @@ def write_text_doc(content): data_doc = { "_id": data_doc_guid, "fields": { - "proto": protofy("textProto"), + "proto": protofy("commonImportProto"), "data": { "Data": '{"doc":{"type":"doc","content":[{"type":"paragraph","content":[{"type":"text","text":"' + content + '"}]}]},"selection":{"type":"text","anchor":1,"head":1}' + '}', "__type": "RichTextField" @@ -158,8 +161,8 @@ def write_text_doc(content): "__type": "Doc" } - db.newDocuments.insert_one(view_doc) - db.newDocuments.insert_one(data_doc) + target_collection.insert_one(view_doc) + target_collection.insert_one(data_doc) return view_doc_guid @@ -209,8 +212,8 @@ def write_image(folder, name): "__type": "Doc" } - db.newDocuments.insert_one(view_doc) - db.newDocuments.insert_one(data_doc) + target_collection.insert_one(view_doc) + target_collection.insert_one(data_doc) return view_doc_guid @@ -231,6 +234,8 @@ def parse_document(file_name: str): for image in os.listdir(dir_path): count += 1 view_guids.append(write_image(pure_name, image)) + copyfile(dir_path + "/" + image, dir_path + + "/" + image.replace(".", "_o.", 1)) os.rename(dir_path + "/" + image, dir_path + "/" + image.replace(".", "_m.", 1)) print(f"extracted {count} images...") @@ -347,6 +352,22 @@ def proxify_guids(guids): return list(map(lambda guid: {"fieldId": guid, "__type": "proxy"}, guids)) +def write_common_proto(): + id = guid() + common_proto = { + "_id": id, + "fields": { + "proto": protofy("collectionProto"), + "title": "Common Import Proto", + }, + "__type": "Doc" + } + + target_collection.insert_one(common_proto) + + return id + + if os.path.exists(dist): shutil.rmtree(dist) while os.path.exists(dist): @@ -354,6 +375,8 @@ while os.path.exists(dist): os.mkdir(dist) mkdir_if_absent(source) +common_proto_id = write_common_proto() + candidates = 0 for file_name in os.listdir(source): if file_name.endswith('.docx'): @@ -372,7 +395,7 @@ parent_guid = write_schema({ }, ["title", "short_description", "original_price"], "data") print("appending parent schema to main workspace...\n") -db.newDocuments.update_one( +target_collection.update_one( {"fields.title": "WS collection 1"}, {"$push": {"fields.data.fields": {"fieldId": parent_guid, "__type": "proxy"}}} ) |