行转列是一种常见的数据处理操作，所以对如何在 SQL 语句中、如何在 pandas 中实现这种行转列做过一些总结。请参考我之前写的博文：
- pandas 行转列一种典型输出报表的解决方法
- 使用 Access 查询设计器轻松构造复杂 SQL 语句
- 使用 Access 查询设计器轻松构造复杂 SQL 语句 (2)
今天介绍在 Power Query (PQ) 实现行转列的操作方法，顺便说是稍微复杂一点的 IF 语句如何编写。
新建一个空查询，导入我已经放在 github 上的示例数据：
代码语言：txt复制let
source = Csv.Document(
Web.Contents("https://raw.githubusercontent.com/stonewm/python-practice-projects/master/pandas sample data/project-listing.csv"),
[Delimiter=",", Columns=10, Encoding=65001, QuoteStyle=QuoteStyle.None])
in
source可以在高级编辑器中拷贝以上代码，或者根据 url 从 Web 导入数据，因为数据是 csv 格式，所以 PQ 用 Csv.Document() 函数来读取数据的内容。
将第一行作为标题，PQ 自动识别数据类型：
因为公式栏和高级编辑器对中文的支持不好，建议将所有步骤改成英文，步骤名称最好不要有空格。
这一步骤完成后，进入高级编辑器，M 脚本代码如下：
代码语言：txt复制let
source = Csv.Document(
Web.Contents("https://raw.githubusercontent.com/stonewm/python-practice-projects/master/pandas sample data/project-listing.csv"),
[Delimiter=",", Columns=10, Encoding=65001, QuoteStyle=QuoteStyle.None]),
firstLineAsTitle = Table.PromoteHeaders(source, [PromoteAllScalars=true]),
changedTitle = Table.TransformColumnTypes(firstLineAsTitle,
{{"ID", Int64.Type},
{"YEAR", Int64.Type},
{"MONTH", Int64.Type},
{"ENTITY", Int64.Type},
{"ACCOUNT", type text},
{"DIRECTION", type text},
{"CURRENCY", type text},
{"PROJ_TYPE", type text},
{"PROJ_CODE", type text},
{"AMOUNT", type number}})
in
changedTitle注意 PQ 将 ACCOUNT 字段识别为 Int64.Type，我改为了 text。
接下来增加条件列来进行行转列的操作，说说规则：
代码语言：txt复制- BEGIN_COST (年初成本： 科目为 140401，期间 < 2018 年
- BEGIN_VAR (年初成本变化额：科目为 140404，期间 < 2018 年
- PER_COST_ADD (期间成本增加：科目为 140401，方向为 S, Year = 2018, Month <= 4
- PER_VAR_ADD (期间公允价值增加：科目为 140404，方向为 S, Year = 2018, Month <= 4
- PER_COST_DECT (期间成本减少：科目为 140401，方向为 H, Year = 2018, Month <= 4
- PER_VAR_ADD (期间公允价值减少：科目为 140404，方向为 H, Year = 2018, Month <= 4增加一个条件列，我们可以看到，PQ 并不能表达 [ACCOUNT] = "140401" and [YEAR] < 2018，对话框中只能根据某一个列来设置条件：
我们需要点击确定后，在公式栏或高级编辑器中编写条件表达式：
选中步骤 addedBeginCost，右键，在后面插入一个步骤
因为步骤的公式类似，可以从上一步骤的公式拷贝过来进行改写：
用类似的方法添加步骤，计算出其他栏位。完成后 M 语言脚本如下：
代码语言：txt复制let
source = Csv.Document(
Web.Contents("https://raw.githubusercontent.com/stonewm/python-practice-projects/master/pandas sample data/project-listing.csv"),
[Delimiter=",", Columns=10, Encoding=65001, QuoteStyle=QuoteStyle.None]),
firstLineAsTitle = Table.PromoteHeaders(source, [PromoteAllScalars=true]),
changedTitle = Table.TransformColumnTypes(firstLineAsTitle,
{{"ID", Int64.Type},
{"YEAR", Int64.Type},
{"MONTH", Int64.Type},
{"ENTITY", Int64.Type},
{"ACCOUNT", type text},
{"DIRECTION", type text},
{"CURRENCY", type text},
{"PROJ_TYPE", type text},
{"PROJ_CODE", type text},
{"AMOUNT", type number}}),
addedBeginCost = Table.AddColumn(changedTitle, "BEGIN_COST", each if [ACCOUNT] = "140401" and [YEAR] < 2018 then [AMOUNT] else 0),
addedBeginVar = Table.AddColumn(addedBeginCost, "BEGIN_VAR", each if [ACCOUNT] = "140404" and [YEAR] < 2018 then [AMOUNT] else 0),
addedPerCostAdd = Table.AddColumn(
addedBeginVar, "PER_COST_ADD", each if [ACCOUNT] = "140401" and [DIRECTION] = "S" and [YEAR] = 2018 and [MONTH] = 4 then [AMOUNT] else 0),
addedPerVarAdd = Table.AddColumn(
addedPerCostAdd, "PER_VAR_ADD", each if [ACCOUNT] = "140404" and [DIRECTION] = "S" and [YEAR] = 2018 and [MONTH] = 4 then [AMOUNT] else 0),
addedPerCostDect = Table.AddColumn(
addedPerVarAdd, "PER_COST_DECT",each if [ACCOUNT] = "140401" and [DIRECTION] = "H" and [YEAR] = 2018 and [MONTH] = 4 then [AMOUNT] else 0),
addedPerVarDect = Table.AddColumn(
addedPerCostDect, "PER_VAR_DECT", each if [ACCOUNT] = "140404" and [DIRECTION] = "H" and [YEAR] = 2018 and [MONTH] = 4 then [AMOUNT] else 0)
in
addedPerVarDect根据 PROJ_CODE 进行分组：
完成之后数据格式如下：
完成之后 M 语言脚本如下：
代码语言：txt复制let
source = Csv.Document(
Web.Contents("https://raw.githubusercontent.com/stonewm/python-practice-projects/master/pandas sample data/project-listing.csv"),
[Delimiter=",", Columns=10, Encoding=65001, QuoteStyle=QuoteStyle.None]),
firstLineAsTitle = Table.PromoteHeaders(source, [PromoteAllScalars=true]),
筛选的行 = Table.SelectRows(firstLineAsTitle, each true),
changedTitle = Table.TransformColumnTypes(筛选的行,
{{"ID", Int64.Type},
{"YEAR", Int64.Type},
{"MONTH", Int64.Type},
{"ENTITY", Int64.Type},
{"ACCOUNT", type text},
{"DIRECTION", type text},
{"CURRENCY", type text},
{"PROJ_TYPE", type text},
{"PROJ_CODE", type text},
{"AMOUNT", type number}}),
addedBeginCost = Table.AddColumn(changedTitle, "BEGIN_COST", each if [ACCOUNT] = "140401" and [YEAR] < 2018 then [AMOUNT] else 0),
addedBeginVar = Table.AddColumn(addedBeginCost, "BEGIN_VAR", each if [ACCOUNT] = "140404" and [YEAR] < 2018 then [AMOUNT] else 0),
addedPerCostAdd = Table.AddColumn(
addedBeginVar, "PER_COST_ADD", each if [ACCOUNT] = "140401" and [DIRECTION] = "S" and [YEAR] = 2018 and [MONTH] <= 4 then [AMOUNT] else 0),
addedPerVarAdd = Table.AddColumn(
addedPerCostAdd, "PER_VAR_ADD", each if [ACCOUNT] = "140404" and [DIRECTION] = "S" and [YEAR] = 2018 and [MONTH] <= 4 then [AMOUNT] else 0),
addedPerCostDect = Table.AddColumn(
addedPerVarAdd, "PER_COST_DECT",each if [ACCOUNT] = "140401" and [DIRECTION] = "H" and [YEAR] = 2018 and [MONTH] <= 4 then [AMOUNT] else 0),
addedPerVarDect = Table.AddColumn(
addedPerCostDect, "PER_VAR_DECT", each if [ACCOUNT] = "140404" and [DIRECTION] = "H" and [YEAR] = 2018 and [MONTH] <= 4 then [AMOUNT] else 0),
grouped = Table.Group(addedPerVarDect, {"PROJ_CODE"}, {
{"BEGIN_COST", each List.Sum([BEGIN_COST]), type number},
{"BEGIN_VAR", each List.Sum([BEGIN_VAR]), type number},
{"PER_COST_ADD", each List.Sum([PER_COST_ADD]), type number},
{"PER_VAR_ADD", each List.Sum([PER_VAR_ADD]), type number},
{"PER_COST_DECT", each List.Sum([PER_COST_DECT]), type number},
{"PER_VAR_DECT", each List.Sum([PER_VAR_DECT]), type number}
}),
sortedLines = Table.Sort(grouped,{{"PROJ_CODE", Order.Ascending}})
in
sortedLines本示例贴了较多代码，作用 1 是结合操作步骤观察和理解 M 代码，作用二是可以直接将代码粘贴到高级编辑器，生成查询，直接学习。
Table.Group() 函数语法：
Table.Group(table as table,
key as any,
aggregatedColumns as list,
optional groupKind as nullable number,
optional comparer as nullable function) as table
