Windows/macOS/Linux下将脚本打包为原生应用的工具汇总
与图形界面应用相比，命令行脚本（Shell、PowerShell、Batch、Python、Node.js 等））相对灵活、可组合性较强，很适合自动化、批处理任务。但对普通用户来说，命令行工具往往过于“硬核”。对普通用户，更合理的交付方式是将脚本封装成原生应用，只需要双击图标即可运行，无需打开终端。
汇总一下Windows/macOS/Linux将命令行脚本打包为原生应用的工具。
ESXi vm-support 支持包分析
--- |
使用 Camelot 合并跨多页的表格
我目前使用 Camelot 从 PDF 文件中提取表格。很多情况下，一个表格会跨越多页，或者包含空白行。这两种情况都会导致 Camelot 生成两个或多个表格（具体取决于空白行的数量或表格跨越的页数）。我想知道在 Camelot 生成多个表格后，是否有办法将这些单独的表格合并成一个表格？或者，是否有办法让 Camelot 忽略分页符和/或空白行？
以下脚本将从给定的 PDF 文件中提取所有表格，并将跨越多页的表格合并。结果将以 CSV 文件格式保存，文件名格式为 page-x-table-y.csv 其中 page-x-table-y 分别代表表格所在的页码和表格在该页码上的索引（因为一页上可能包含多个表格）。对于合并的表格，文件名中的 x 代表表格起始页码。
请注意，此脚本仅提取带有边框的表格，不适用于其他类型的表格——这意味着它使用的是 Camelot 的 lattice 模式而非 stream 模式 。因此，如有必要，您可能需要根据自身需求进行调整。
如果 PDF 文件中没有表格，则不会生成任何文件！
将以下代码保存到名为 pdf-extract-tables.py 的文件中，然后像这样调用它python /path/to/pdf-extract-tables.py -p /path/to/file.pdf -t 15
实现神奇效果的代码：import argparse
import warnings
import camelot
import csv
import os
def get_continued_tables(tables, threshold):
continued_tables = {}
previous_table = False
group_counter = 0
# typical height of a pdf is 842 points and bottom margins are anywhere between 56 and 85 points
# therefore, accounting for margins, 792
page_height = 792
# iterate over the tables
for i, table in enumerate(tables):
# if a previous table exists (remember, we start with this as false)
# and the previous table was on the previous page
# and the number of columns of both tables is the same
if previous_table and table.page == previous_table.page + 1 and len(table.cols) == len(previous_table.cols):
# get the bottom coordinate of the previous table
# note that for pdfs the origin (0, 0) typically starts from the bottom-left corner of the page,
# with the y-coordinate increasing as you move upwards
# this is why for {x0, y0, x1, y1} we need the y0 as the bottom
previous_table_bottom = previous_table._bbox[1]
# get the top coordinate of the current table
# for {x0, y0, x1, y1} we need the y1 as the top
current_table_top = table._bbox[3]
# if the previous table ends in the last 15% of the page and the current table starts in the first 15% of the page
if previous_table_bottom < (threshold / 100) * page_height and current_table_top > (1 - threshold / 100) * page_height:
# if we don't have started this group of tables
if (continued_tables.get(group_counter) is None):
# start by adding the first table
continued_tables[group_counter] = [previous_table]
# add any of the sunsequent tables to the group
continued_tables[group_counter].append(table)
# if this is not a continuation of the previous table
else:
# increment the group number
group_counter += 1;
# if this is not a continuation of the previous table
else:
# increment the group number
group_counter += 1;
# the current table becomes the previous table for the next iteration
previous_table = table
# transform the dictionary into an array of arrays
continued_tables = [value for value in continued_tables.values()]
# return the combined tables
return continued_tables
def main():
class NewlineFormatter(argparse.RawDescriptionHelpFormatter):
def _split_lines(self, text, width):
return text.splitlines()
# create argument parser
parser = argparse.ArgumentParser(description = 'Returns an array of tables that should be grouped, also as an array.\nThe tables that should be grouped represent tables that span over multiple pages.', formatter_class = NewlineFormatter)
parser.add_argument('--path', '-p', type = str, metavar = '', required = True, help = 'path to the PDF file containing tables')
parser.add_argument('--threshold', '-t', type = int, metavar = '', default = 15, help = 'if the table on previous page ends in the last x%% of the page and\nthe table on the next page starts in the first x%% of the page,\ntables will be considered as spanning over those pages.\nDefault is 15')
# parse command-line arguments
args = parser.parse_args()
# suppress warning about no tables found
warnings.filterwarnings('ignore', message = 'No tables found on page-*')
# extract tables
tables = camelot.read_pdf(args.path, flavor = 'lattice', pages = 'all')
# get continued tables
continued_tables = get_continued_tables(tables, args.threshold);
# get the name of the PDF file we are processing (without the extension)
pdf_file_name = os.path.basename(args.path)
# the path where we're writing the file to
# (same place where we processed the file from)
file_path = os.path.dirname(args.path)
written = []
# the number of rows to be written at once to the CSV
# (writing rows one by one is less efficient)
batch_size = 1000
# iterate over found tables
for i, table in enumerate(tables):
# if table was already written as part of a group
if table in written: continue
# check if the current table is a continued table
is_continued = any(table in sublist for sublist in continued_tables)
# define the filename for the CSV file (the PDF file's name suffixed with the page where the table was found and the table's index on that page)
# (the naming is the same as what camelot does when called from the CLI)
file_name = f"{pdf_file_name}-page-{table.parsing_report['page']}-table-{table.parsing_report['order']}.csv"
# open the CSV file for writing
with open(file_path + '/' + file_name, 'w', newline = '') as csv_file:
# create a CSV writer
csv_writer = csv.writer(csv_file)
batch = []
# iterate over the rows in the table
for row in table.data:
# add row to batch
batch.append(row)
# if we gathered enough rows, or if we're at the last row
if len(batch) >= batch_size or row == table.data[-1]:
# write batch to the CSV file
csv_writer.writerows(batch)
# clear the batch
batch = []
# if there are any records left to be written
# write batch to the CSV file
if batch: csv_writer.writerows(batch)
# if the current table is a continued table, write all subsequent continued tables to the same CSV file
if is_continued:
# get the index of the group in "continued_tables" associated with the current table
group_index = next(index for index, sublist in enumerate(continued_tables) if table in sublist)
# iterate over the tables in said group and append their data to the same CSV file
for continued_table in continued_tables[group_index]:
# skip writing the current table as it's already written
if continued_table == table or continued_table in written: continue
# write the data of the continued table to the same CSV file
with open(file_path + '/' + file_name, 'a', newline = '') as csv_file:
# create a CSV writer
csv_writer = csv.writer(csv_file)
batch = []
# iterate over the rows in the table
for row in continued_table.data:
# add row to batch
batch.append(row)
# if we gathered enough rows, or if we're at the last row
if len(batch) >= batch_size or row == table.data[-1]:
# write batch to the CSV file
csv_writer.writerows(batch)
# clear the batch
batch = []
# if there are any records left to be written
# write batch to the CSV file
if batch: csv_writer.writerows(batch)
# keep track of written tables so that are not written again in the main iteration
written.append(continued_table)
if __name__ == "__main__":
main()
Windows 端口未被占用却提示“端口冲突”的排查与解决
¶1. 问题现象
在 Windows 系统中启动应用（如 Tomcat、Node.js、Spring Boot 等）时，系统抛出 Address already in use (端口已被占用) 错误。然而，使用常规命令 netstat -ano | findstr ":8080" 查询时，却**没有任何进程（PID）**在监听该端口。
¶2. 根本原因（Why）
该现象由 Windows Hyper-V / WSL2 / Docker Desktop 与 WinNAT（网络地址转换服务） 的底层机制冲突导致：
- 系统保留机制：WinNAT 服务在系统启动时，会随机、成批地向系统内核申请保留多段 TCP 端口（通常每组 100 个），用于虚拟机或容器的内部通信。
- 默认范围过低：Windows 默认的动态端口起始位置偏低，导致系统在随机霸占端口时，经常误将开发者常用的核心端口（如
8080,3000,9000）锁死。 - 特征：这些被系统保留的端口在
netstat中对普通应用显示为“空闲”，但普通应用无权绑定。且每次重启电脑后，被保留的端口区间会随机发生变化。
¶3. 排查方法
打开具有 管理员权限 的命令提示符（CMD）或 PowerShell，运行以下命令查看系统当前排他的端口范围：netsh int ipv4 show excludedportrange protocol=tcp
结果分析：
若输出的表格中，目标端口（例如 8080）落在了某一组 开始端口 和 结束端口 的区间内，即可证实该端口已被系统强制保留。
¶4. 彻底解决方案（One-Time Fix）
为了彻底解决并防止系统未来继续随机侵占低位常用开发端口，最有效的办法是将 Windows 的动态端口起始位置移至国际标准高位段（49152）。
请在 管理员权限 的终端中，严格按照以下步骤执行：
¶步骤一：暂停网络转换服务
因为部分端口当前正被系统占用，直接修改可能会触发“拒绝访问”或“文件正被使用”的错误。需要先暂停 WinNAT 服务：net stop winnat
¶步骤二：重置动态端口范围
将 TCP 和 UDP 的动态端口起点强制修改为 49152（共分配 16383 个端口）。修改后，系统服务将只在 49152~65535 之间随机保留端口，永远不会再触碰 8080 等低位端口。netsh int ipv4 set dynamicport tcp start=49152 num=16383
netsh int ipv4 set dynamicport udp start=49152 num=16383
¶步骤三：重新启动网络服务
net start winnat |
¶步骤四：重启计算机
执行完上述操作后，必须重启电脑使配置在系统内核中彻底生效。
¶5. 验证结果
电脑重启后，再次以管理员身份运行排查命令：netsh int ipv4 show excludedportrange protocol=tcp
此时会发现诸如 8023-8122 这一类处于低位段（10000以下）的保留区间已全部移至 49152 之后，8080 端口成功释放，应用可正常启动。