在 IMPALA/HIVE 中添加带有 SELECT 的新列后，旧表数据变为 NULL

Question

我正在尝试使用 (SELECT,JOIN) 查询将数据添加到 Impala 中的新列，一旦我将数据添加到新列，我就会丢失所有其他列的数据（它们变为 NULL）。

在这里我创建了第一个表：

CREATE TABLE mng_exp.KPI_LATENCE_JOUR
(
    CODEINSEE       INT,
    IMEI            BIGINT,
    SEMAINE         INT,
    MOYENNE_LATENCE INT,
    MAXIMUM_LATENCE INT,
    MINIMUM_LATENCE INT
)

我在表格中添加数据：

INSERT INTO mng_exp.KPI_LATENCE_JOUR (CODEINSEE,IMEI, SEMAINE, MOYENNE_LATENCE,MAXIMUM_LATENCE,MINIMUM_LATENCE,TRANCHE_DE_LATENCE) 
SELECT codeinsee, device_dim__imei as IMEI,weekofyear(jour) as SEMAINE, cast(round(avg(rtt_avg_ms)) as integer) as MOYENNE_LATENCE, 
cast(round(avg(rtt_max_ms)) as integer) as MAXIMUM_LATENCE, cast(round(avg(rtt_min_ms)) as integer) as MINIMUM_LATENCE , 
CASE WHEN ( round(avg(rtt_avg_ms)) > 0 and round(avg(rtt_avg_ms)) <= 10 ) THEN 0 
WHEN ( round(avg(rtt_avg_ms)) > 10 and round(avg(rtt_avg_ms)) <= 20 ) THEN 1 
WHEN ( round(avg(rtt_avg_ms)) > 20 and round(avg(rtt_avg_ms)) <= 30 ) THEN 2 
WHEN ( round(avg(rtt_avg_ms)) > 30 ) THEN 3 END AS Tranche_de_latence

FROM mscore.mscore where operateur = 'BT_HZ' and year(jour) = 2019 group by device_dim__imei,weekofyear(jour),codeinsee 



# I  Add a new column

ALTER TABLE mng_exp.kpi_latence_jour ADD COLUMNS (srv_id BIGINT)

#Here data is good and new column srv_id is NULL 

我将数据添加到新列：

INSERT INTO mng_exp.KPI_LATENCE_jour (srv_id) 
    SELECT CAST(dng_fai_cli_eqt_iad.srv_id AS BIGINT)
    FROM msf_exploratoire.dng_fai_cli_eqt_iad
    INNER JOIN mng_exp.kpi_latence_jour ON (dng_fai_cli_eqt_iad.num_serie = kpi_latence_jour.imei);

问题出在：srv_id 正常，旧列变为 NULL。

我没有查询错误，但我丢失了所有旧数据

Answer 1

您只插入一列。使用 INSERT OVERWRITE 并添加所有其他列：

INSERT OVERWRITE TABLE mng_exp.KPI_LATENCE_jour (CODEINSEE,IMEI, SEMAINE, MOYENNE_LATENCE,MAXIMUM_LATENCE,MINIMUM_LATENCE,TRANCHE_DE_LATENCE,srv_id) 
    SELECT b.CODEINSEE,
           b.IMEI,
           b.SEMAINE,
           b.MOYENNE_LATENCE,
           b.MAXIMUM_LATENCE,
           b.MINIMUM_LATENCE,
           b.TRANCHE_DE_LATENCE,
           CAST(a.srv_id AS BIGINT) srv_id
      FROM msf_exploratoire.dng_fai_cli_eqt_iad a
           INNER JOIN mng_exp.kpi_latence_jour  b ON (a.num_serie = b.imei)
      ;

Answer 2

您确定丢失了所有旧数据还是执行：

select * from mng_exp.KPI_LATENCE_JOUR

你还会看到：

1. 第一组行（srv_id 为空的行）；
2. 第二组行 - 唯一填充的列是 srv_id 的行？

您想要更新第一组行的 SRV_ID。

您可以在 imapala 中查看有关更新的更多详细信息，here。