Мне очень помог ответ Криса, но я хотел запустить его из SQL Server, используя T-SQL (а не CLR), поэтому я преобразовал его код в код T-SQL. Но затем я сделал еще один шаг, обернув все в хранимую процедуру, которая сделала следующее:
- использовать массовую вставку для первоначального импорта файла CSV
- очистить строки, используя код Криса
- вернуть результаты в виде таблицы
Для своих нужд я дополнительно очистил строки, удалив кавычки вокруг значений и преобразовав две двойные кавычки в одну двойную кавычку (я думаю, это правильный метод).
CREATE PROCEDURE SSP_CSVToTable
-- Add the parameters for the stored procedure here
@InputFile nvarchar(4000)
, @FirstLine int
AS
BEGIN
-- SET NOCOUNT ON added to prevent extra result sets from
-- interfering with SELECT statements.
SET NOCOUNT ON;
--convert the CSV file to a table
--clean up the lines so that commas are handles correctly
DECLARE @sql nvarchar(4000)
DECLARE @PH1 nvarchar(50)
DECLARE @LINECOUNT int -- This will also serve as a primary key
DECLARE @CURLINE int
DECLARE @Line nvarchar(4000)
DECLARE @starti int
DECLARE @endi int
DECLARE @FieldTerminatorFound bit
DECLARE @backChar nvarchar(4000)
DECLARE @quoteCount int
DECLARE @source nvarchar(4000)
DECLARE @COLCOUNT int
DECLARE @CURCOL int
DECLARE @ColVal nvarchar(4000)
-- new delimiter
SET @PH1 = '†'
-- create single column table to hold each line of file
CREATE TABLE [#CSVLine]([line] nvarchar(4000))
-- bulk insert into temp table
-- cannot use variable path with bulk insert
-- so we must run using dynamic sql
SET @Sql = 'BULK INSERT #CSVLine
FROM ''' + @InputFile + '''
WITH
(
FIRSTROW=' + CAST(@FirstLine as varchar) + ',
FIELDTERMINATOR = ''\n'',
ROWTERMINATOR = ''\n''
)'
-- run dynamic statement to populate temp table
EXEC(@sql)
-- get number of lines in table
SET @LINECOUNT = @@ROWCOUNT
-- add identity column to table so that we can loop through it
ALTER TABLE [#CSVLine] ADD [RowId] [int] IDENTITY(1,1) NOT NULL
IF @LINECOUNT > 0
BEGIN
-- cycle through each line, cleaning each line
SET @CURLINE = 1
WHILE @CURLINE <= @LINECOUNT
BEGIN
-- get current line
SELECT @line = line
FROM #CSVLine
WHERE [RowId] = @CURLINE
-- Replace commas with our custom-made delimiter
SET @Line = REPLACE(@Line, ',', @PH1)
-- Find a quoted part of the line, which could legitimately contain commas.
-- In that case we will need to identify the quoted section and swap commas back in for our custom placeholder.
SET @starti = CHARINDEX(@PH1 + '"' ,@Line, 0)
If CHARINDEX('"', @Line, 0) = 0 SET @starti = 0
-- loop through quoted fields
WHILE @starti > 0
BEGIN
SET @FieldTerminatorFound = 0
-- Find end quote token (originally a ",)
SET @endi = CHARINDEX('"' + @PH1, @Line, @starti) -- sLine.IndexOf("""" & PH1, starti)
IF @endi < 1
BEGIN
SET @FieldTerminatorFound = 1
If @endi < 1 SET @endi = LEN(@Line) - 1
END
WHILE @FieldTerminatorFound = 0
BEGIN
-- Find any more quotes that are part of that sequence, if any
SET @backChar = '"' -- thats one quote
SET @quoteCount = 0
WHILE @backChar = '"'
BEGIN
SET @quoteCount = @quoteCount + 1
SET @backChar = SUBSTRING(@Line, @endi-@quoteCount, 1) -- sLine.Chars(endi - quoteCount)
END
IF (@quoteCount % 2) = 1
BEGIN
-- odd number of quotes. real field terminator
SET @FieldTerminatorFound = 1
END
ELSE
BEGIN
-- keep looking
SET @endi = CHARINDEX('"' + @PH1, @Line, @endi + 1) -- sLine.IndexOf("""" & PH1, endi + 1)
END
END
-- Grab the quoted field from the line, now that we have the start and ending indices
SET @source = SUBSTRING(@Line, @starti + LEN(@PH1), @endi - @starti - LEN(@PH1) + 1)
-- sLine.Substring(starti + PH1.Length, endi - starti - PH1.Length + 1)
-- And swap the commas back in
SET @Line = REPLACE(@Line, @source, REPLACE(@source, @PH1, ','))
--sLine.Replace(source, source.Replace(PH1, ","))
-- Find the next quoted field
-- If endi >= line.Length - 1 Then endi = line.Length 'During the swap, the length of line shrinks so an endi value at the end of the line will fail
SET @starti = CHARINDEX(@PH1 + '"', @Line, @starti + LEN(@PH1))
--sLine.IndexOf(PH1 & """", starti + PH1.Length)
END
-- get table based on current line
IF OBJECT_ID('tempdb..#Line') IS NOT NULL
DROP TABLE #Line
-- converts a delimited list into a table
SELECT *
INTO #Line
FROM dbo.iter_charlist_to_table(@Line,@PH1)
-- get number of columns in line
SET @COLCOUNT = @@ROWCOUNT
-- dynamically create CSV temp table to hold CSV columns and lines
-- only need to create once
IF OBJECT_ID('tempdb..#CSV') IS NULL
BEGIN
-- create initial structure of CSV table
CREATE TABLE [#CSV]([Col1] nvarchar(100))
-- dynamically add a column for each column found in the first line
SET @CURCOL = 1
WHILE @CURCOL <= @COLCOUNT
BEGIN
-- first column already exists, don't need to add
IF @CURCOL > 1
BEGIN
-- add field
SET @sql = 'ALTER TABLE [#CSV] ADD [Col' + Cast(@CURCOL as varchar) + '] nvarchar(100)'
--print @sql
-- this adds the fields to the temp table
EXEC(@sql)
END
-- go to next column
SET @CURCOL = @CURCOL + 1
END
END
-- build dynamic sql to insert current line into CSV table
SET @sql = 'INSERT INTO [#CSV] VALUES('
-- loop through line table, dynamically adding each column value
SET @CURCOL = 1
WHILE @CURCOL <= @COLCOUNT
BEGIN
-- get current column
Select @ColVal = str
From #Line
Where listpos = @CURCOL
IF LEN(@ColVal) > 0
BEGIN
-- remove quotes from beginning if exist
IF LEFT(@ColVal,1) = '"'
SET @ColVal = RIGHT(@ColVal, LEN(@ColVal) - 1)
-- remove quotes from end if exist
IF RIGHT(@ColVal,1) = '"'
SET @ColVal = LEFT(@ColVal, LEN(@ColVal) - 1)
END
-- write column value
-- make value sql safe by replacing single quotes with two single quotes
-- also, replace two double quotes with a single double quote
SET @sql = @sql + '''' + REPLACE(REPLACE(@ColVal, '''',''''''), '""', '"') + ''''
-- add comma separater except for the last record
IF @CURCOL <> @COLCOUNT
SET @sql = @sql + ','
-- go to next column
SET @CURCOL = @CURCOL + 1
END
-- close sql statement
SET @sql = @sql + ')'
--print @sql
-- run sql to add line to table
EXEC(@sql)
-- move to next line
SET @CURLINE = @CURLINE + 1
END
END
-- return CSV table
SELECT * FROM [#CSV]
END
GO
Хранимая процедура использует эту вспомогательную функцию, которая анализирует строку в таблице (спасибо Erland Sommarskog!):
CREATE FUNCTION [dbo].[iter_charlist_to_table]
(@list ntext,
@delimiter nchar(1) = N',')
RETURNS @tbl TABLE (listpos int IDENTITY(1, 1) NOT NULL,
str varchar(4000),
nstr nvarchar(2000)) AS
BEGIN
DECLARE @pos int,
@textpos int,
@chunklen smallint,
@tmpstr nvarchar(4000),
@leftover nvarchar(4000),
@tmpval nvarchar(4000)
SET @textpos = 1
SET @leftover = ''
WHILE @textpos <= datalength(@list) / 2
BEGIN
SET @chunklen = 4000 - datalength(@leftover) / 2
SET @tmpstr = @leftover + substring(@list, @textpos, @chunklen)
SET @textpos = @textpos + @chunklen
SET @pos = charindex(@delimiter, @tmpstr)
WHILE @pos > 0
BEGIN
SET @tmpval = ltrim(rtrim(left(@tmpstr, @pos - 1)))
INSERT @tbl (str, nstr) VALUES(@tmpval, @tmpval)
SET @tmpstr = substring(@tmpstr, @pos + 1, len(@tmpstr))
SET @pos = charindex(@delimiter, @tmpstr)
END
SET @leftover = @tmpstr
END
INSERT @tbl(str, nstr) VALUES (ltrim(rtrim(@leftover)), ltrim(rtrim(@leftover)))
RETURN
END
Вот как я это называю из T-SQL. В этом случае я вставляю результаты в временную таблицу, поэтому сначала создаю временную таблицу:
-- create temp table for file import
CREATE TABLE #temp
(
CustomerCode nvarchar(100) NULL,
Name nvarchar(100) NULL,
[Address] nvarchar(100) NULL,
City nvarchar(100) NULL,
[State] nvarchar(100) NULL,
Zip nvarchar(100) NULL,
OrderNumber nvarchar(100) NULL,
TimeWindow nvarchar(100) NULL,
OrderType nvarchar(100) NULL,
Duration nvarchar(100) NULL,
[Weight] nvarchar(100) NULL,
Volume nvarchar(100) NULL
)
-- convert the CSV file into a table
INSERT #temp
EXEC [dbo].[SSP_CSVToTable]
@InputFile = @FileLocation
,@FirstLine = @FirstImportRow
Я не очень тестировал производительность, но она хорошо работает для того, что мне нужно - импортировать файлы CSV с менее чем 1000 строк. Тем не менее, он может подавиться очень большими файлами.
Надеюсь, кто-то еще найдет это полезным.
Ура!