• новости
• статьи
• блоги
• файлы
• пресс-релизы

статей часть 1 и часть 2, в которых предложено решение задачи визуализации иерархической структуры или ее части средствами SQL запросов на примере MySQL и SQLite

with recursive 
Mapping as (
	select 
		id 					as node_id, 
		parent_directory_id as parent_node_id,
		name 				as node_name
	from Files
),

RootNodes as (
	select node_id as root_node_id
	from Mapping
	where 						-- Exactly one line below should be uncommented
		-- parent_node_id is null	-- Uncomment to build from root(s)
		node_id in (3, 10, 17)	-- Uncomment to add node_id(s) into the brackets
),

Levels as (
	select
		node_id,
		parent_node_id,
		node_name,
		cast(parent_node_id as varchar) as parents,
		cast(node_name as varchar) as full_path,
		0 as node_level
	from
		Mapping
		inner join RootNodes on node_id = root_node_id
		
	union
	
	select 
		Mapping.node_id, 
		Mapping.parent_node_id,
		Mapping.node_name,
		concat(coalesce(concat(prev.parents, '-'), ''), cast(Mapping.parent_node_id as varchar)),
		concat_ws(' ', prev.full_path, Mapping.node_name),
		prev.node_level + 1
	from 
		Levels as prev
		inner join Mapping on Mapping.parent_node_id = prev.node_id
),

Branches as (
	select
		node_id,
		parent_node_id,
		node_name,
		parents,
		full_path,
		node_level,
		case
			when root_node_id is null then
				case
					when node_id = last_value(node_id) over WindowByParents then '└── '
					else '├── '
				end
			else ''
		end as node_branch,
		case
			when root_node_id is null then
				case
					when node_id = last_value(node_id) over WindowByParents then '    '
					else '│   '
				end
			else ''
		end as branch_through
	from
		Levels
		left join RootNodes on node_id = root_node_id
	window WindowByParents as (
		partition by parents
		order by node_name
		rows between current row and unbounded following
		)
	order by full_path
),

Tree as (
	select
		node_id,
		parent_node_id,
		node_name,
		parents,
		full_path,
		node_level,
		node_branch,
		cast(branch_through as varchar) as all_through
	from
		Branches
		inner join RootNodes on node_id = root_node_id
		
	union
	
	select 
		Branches.node_id,
		Branches.parent_node_id,
		Branches.node_name,
		Branches.parents,
		Branches.full_path,
		Branches.node_level,
		Branches.node_branch,
		concat(prev.all_through, Branches.branch_through)
	from 
		Tree as prev
		inner join Branches on Branches.parent_node_id = prev.node_id
),

FineTree as (
	select
		tr.node_id,
		tr.parent_node_id,
		tr.node_name,
		tr.parents,
		tr.full_path,
		tr.node_level,
		concat(coalesce(parent.all_through, ''), tr.node_branch, tr.node_name) as fine_tree
	from
		Tree as tr
		left join Tree as parent on
			parent.node_id = tr.parent_node_id
	order by tr.full_path
)

select fine_tree, node_id from FineTree
;

#
# Generate SQL-script for creation of hierarchical table by zip-archive structure
#
from zipfile import ZipFile
from itertools import count


def file_size_b(file_size):
    """ Returns file size string in XBytes """
    for b in ["B", "KB", "MB", "GB", "TB"]:
        if file_size < 1024:
            break
        file_size /= 1024
    return f"{round(file_size)} {b}"


def gen_table(zip_file, table_name='ZipArchive', chunk_size=10000, out_extension='sql'):
    """
    by iqu 2024-04-28
    params:
        zip_file - zip archive full path
        table_name - table to be created
        chunk_size - limit values() for each insert into part
        out extension - replaces zip archive file extension
    ->  None, creates file with SQL script.
        Create table columns:
        id, name, parent_id, file_size - obvious,
        bytes - string with file size generated by file_size_b()
    """
    def gen_create_table(file):
        print(f"drop table if exists {table_name};", file=file)
        print(f"create table {table_name} (id int, name varchar(255), parent_id int, file_size int, bytes varchar(16))"
              , file=file)

    def gen_insert(file):
        print(f";\ninsert into {table_name} (id, name, parent_id, file_size, bytes) values", file=file)

    out_file = ".".join(zip_file.split(".")[:-1] + [out_extension])

    cnt = count()
    parents = ['NULL']

    with open(out_file, mode='w') as of:
        gen_create_table(of)

        with ZipFile(zip_file) as zf:
            for zi in zf.infolist():
                zi_id = cnt.__next__()
                if zi_id % chunk_size == 0:
                    gen_insert(of)
                    delimiter = ''
                else:
                    delimiter = ','

                level = zi.filename.count("/") - zi.is_dir()
                name = zi.filename.split("/")[level]
                file_size = -1 if zi.is_dir() else zi.file_size
                file_size_s = 'DIR' if zi.is_dir() else file_size_b(zi.file_size)
                if zi.is_dir():
                    if len(parents) < level + 2:
                        parents.append(f"{zi_id}")
                    else:
                        parents[level + 1] = f"{zi_id}"

                print(f"{delimiter}({zi_id}, '{name}', {parents[level]}, {file_size}, '{file_size_s}')", file=of)
            print(';', file=of)


gen_table(r"C:\TEMP\jdk-master.zip")    # Sample archive https://github.com/openjdk/jdk/archive/refs/heads/master.zip

Объяснять код в деталях не буду, это совсем не по теме статьи. Существенно, что скрипт разбивается на части длиной максимум chunk_size (10 000 по-умолчанию) в каждом блоке values()

Кроме ИД, имени и ссылки на родителя, каждая запись содержит поле file_size с длиной в байтах (-1 для папок) и поле bytes с длиной, преобразованной к строке в байтах, килобайтах и т.д. ('DIR' для папок)