問題描述
Pyspark DataFrameWriter 類有一個(gè) jdbc 函數(shù) 用于將數(shù)據(jù)幀寫入 sql.這個(gè)函數(shù)有一個(gè) --ignore 選項(xiàng)，文檔說:
The Pyspark DataFrameWriter class has a jdbc function for writing a dataframe to sql. This function has an --ignore option that the documentation says will:
如果數(shù)據(jù)已經(jīng)存在，則靜默忽略此操作.
Silently ignore this operation if data already exists.
但是它會(huì)忽略整個(gè)事務(wù)，還是只會(huì)忽略插入重復(fù)的行?如果我將 --ignore 與 --append 標(biāo)志結(jié)合起來會(huì)怎樣?行為會(huì)改變嗎?
But will it ignore the entire transaction, or will it only ignore inserting the rows that are duplicates? What if I were to combine --ignore with the --append flag? Would the behavior change?
推薦答案
mode("ingore") 如果表(或另一個(gè)接收器)已經(jīng)存在并且無法組合寫入模式，則只是 NOOP.如果您正在尋找諸如 INSERT IGNORE 或 INSERT INTO ... WHERE NOT EXISTS ... 之類的內(nèi)容，則必須手動(dòng)執(zhí)行，例如使用 mapPartitions.
mode("ingore") is just NOOP if table (or another sink) already exists and writing modes cannot be combined. If you're looking for something like INSERT IGNORE or INSERT INTO ... WHERE NOT EXISTS ... you'll have to do it manually, for example with mapPartitions.
這篇關(guān)于Pyspark DataFrameWriter jdbc 函數(shù)的 ignore 選項(xiàng)是忽略整個(gè)事務(wù)還是只是有問題的行?的文章就介紹到這了，希望我們推薦的答案對(duì)大家有所幫助，也希望大家多多支持html5模板網(wǎng)！