优化Cypher查询的性能
这是我上一个问题的后续问题：Correct way to create json from neo4j rest API。
我能够通过@Michael Hunger提供给我的帮助，以我所要求的格式获得JSON。以下是代码：
MATCH (mainCap:Capability{SectorID:"PNU"})
OPTIONAL MATCH (mainCap)-[r1:CapabilityAdopts_Capability]->(adoptsCap1)
OPTIONAL MATCH (adoptsCap1)-[r2:KPI_Capability]->(capKPI)
OPTIONAL MATCH (adoptsCap1)-[r4:Capability_Capability]->(cap1)
OPTIONAL MATCH (cap1)<-[r5:Capability_Capability]-(isPart1)
OPTIONAL MATCH (cap1)-[r6:KPI_Capability]->(KPI1)
OPTIONAL MATCH (cap1)-[r7:Capability_Capability]->(cap2)
OPTIONAL MATCH (cap2)<-[r8:Capability_Capability]-(isPart2)
OPTIONAL MATCH (cap2)-[r9:KPI_Capability]->(KPI2)
OPTIONAL MATCH (cap2)-[r10:Capability_Capability]->(cap3)
OPTIONAL MATCH (cap3)<-[r11:Capability_Capability]-(isPart3)
OPTIONAL MATCH (cap3)-[r12:KPI_Capability]->(KPI3)
OPTIONAL MATCH (mainCap)<-[r3:Capability_Capability]-(isPart:Capability)
with
mainCap as mainCap,
isPart as isPart,
adoptsCap1 as adoptsCap1,
cap1 as cap1,
cap2 as cap2,
cap3 as cap3,
KPI2 as KPI2,
isPart2 as isPart2,
KPI1 as KPI1,
isPart1 as isPart1,
capKPI as capKPI,
CASE WHEN KPI3 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: KPI3.ExternalID,
type: labels(KPI3)[0],
name: KPI3.Name,
description: KPI3.Description,
sectorID: KPI3.SectorID,
CCO: KPI3.CCO,
CFO: KPI3.CFO,
SVPHR: KPI3.SVPHR,
area: KPI3.Area,
CEO: KPI3.CEO,
CIO: KPI3.CIO,
COO: KPI3.COO
}
)
ELSE
[]
END as cpbltyIsMsrdByKpiPttrn3,
CASE WHEN isPart3 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: isPart3.ExternalID,
type: labels(isPart3)[0]
}
)
ELSE
[]
END as isPartOfCapability3
with
mainCap as mainCap,
isPart as isPart,
adoptsCap1 as adoptsCap1,
cap1 as cap1,
cap2 as cap2,
KPI1 as KPI1,
isPart1 as isPart1,
capKPI as capKPI,
CASE WHEN cap3 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: cap3.ExternalID,
type: labels(cap3)[0],
name: cap3.Name,
description: cap3.Overview,
sectorID: cap3.SectorID,
cpbltyIsMsrdByKpiPttrn:cpbltyIsMsrdByKpiPttrn3,
isPartOfCapability:isPartOfCapability3
}
)
ELSE
[]
END as consistsOfCapability3,
CASE WHEN KPI2 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: KPI2.ExternalID,
type: labels(KPI2)[0],
name: KPI2.Name,
description: KPI2.Description,
sectorID: KPI2.SectorID,
CCO: KPI2.CCO,
CFO: KPI2.CFO,
SVPHR: KPI2.SVPHR,
area: KPI2.Area,
CEO: KPI2.CEO,
CIO: KPI2.CIO,
COO: KPI2.COO
}
)
ELSE
[]
END as cpbltyIsMsrdByKpiPttrn2,
CASE WHEN isPart2 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: isPart2.ExternalID,
type: labels(isPart2)[0]
}
)
ELSE
[]
END as isPartOfCapability2
with mainCap as mainCap,
adoptsCap1 as adoptsCap1,
isPart as isPart,
cap1 as cap1,
capKPI as capKPI,
CASE WHEN cap2 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: cap2.ExternalID,
type: labels(cap2)[0],
name: cap2.Name,
description: cap2.Overview,
sectorID: cap2.SectorID,
cpbltyIsMsrdByKpiPttrn:cpbltyIsMsrdByKpiPttrn2,
isPartOfCapability:isPartOfCapability2,
consistsOfCapability:consistsOfCapability3
}
)
ELSE
[]
END as consistsOfCapability2,
CASE WHEN KPI1 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: KPI1.ExternalID,
type: labels(KPI1)[0],
name: KPI1.Name,
description: KPI1.Description,
sectorID: KPI1.SectorID,
CCO: KPI1.CCO,
CFO: KPI1.CFO,
SVPHR: KPI1.SVPHR,
area: KPI1.Area,
CEO: KPI1.CEO,
CIO: KPI1.CIO,
COO: KPI1.COO
}
)
ELSE
[]
END as cpbltyIsMsrdByKpiPttrn1,
CASE WHEN isPart1 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: isPart1.ExternalID,
type: labels(isPart1)[0]
}
)
ELSE
[]
END as isPartOfCapability1
with mainCap as mainCap,
adoptsCap1 as adoptsCap1,
isPart as isPart,
CASE WHEN cap1 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: cap1.ExternalID,
type: labels(cap1)[0],
name: cap1.Name,
description: cap1.Overview,
sectorID: cap1.SectorID,
cpbltyIsMsrdByKpiPttrn:cpbltyIsMsrdByKpiPttrn1,
isPartOfCapability:isPartOfCapability1,
consistsOfCapability:consistsOfCapability2
}
)
ELSE
[]
END as consistsOfCapability1,
CASE WHEN capKPI IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: capKPI.ExternalID,
type: labels(capKPI)[0],
name: capKPI.Name,
description: capKPI.Description,
sectorID: capKPI.SectorID,
CCO: capKPI.CCO,
CFO: capKPI.CFO,
SVPHR: capKPI.SVPHR,
area: capKPI.Area,
CEO: capKPI.CEO,
CIO: capKPI.CIO,
COO: capKPI.COO
}
)
ELSE
[]
END as cpbltyIsMsrdByKpiPttrn
return {
_id:mainCap.ExternalID,
type:labels(mainCap)[0],
name:mainCap.Name,
adoptsCapability:
CASE WHEN adoptsCap1 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: adoptsCap1.ExternalID,
type: labels(adoptsCap1)[0],
name: adoptsCap1.Name,
description: adoptsCap1.Overview,
sectorID: adoptsCap1.SectorID,
cpbltyIsMsrdByKpiPttrn:cpbltyIsMsrdByKpiPttrn,
consistsOfCapability:consistsOfCapability1
}
)
ELSE
[]
END,
isPartOfCapability:
CASE WHEN isPart IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: isPart.ExternalID,
type: labels(isPart)[0]
}
)
ELSE
[]
END
}as `col2`
以上查询采用 4149 ms 来提供结果。我尝试使用distinct进行优化。通过： http://thought-bytes.blogspot.com/2013/01/optimizing-neo4j-cypher-queries.html。 Optimize Neo4j Cypher query 更相关：Optimizing Cypher queries with lots of optional relationships
问题：是否可以优化并将时间缩短到约2秒左右？我有一个angularjs UI，它有一个菜单绑定来自neo的数据。它目前需要4.4秒才能从neo获取数据或者这是我可以从这个查询中获得的最大性能吗？
以下是我的执行计划：
帮助表示感谢。谢谢
以下是解决方案：
重新格式化查询后，性能大幅提升。基数导致性能滞后。 从4149毫秒下降到400毫秒：
match (mainCap:Capability{SectorID:"PNU"})
OPTIONAL MATCH (mainCap:Capability)-[r1:CapabilityAdopts_Capability]->(adoptsCap1)
OPTIONAL MATCH (adoptsCap1)-[r2:KPI_Capability]->(capKPI)
OPTIONAL MATCH (adoptsCap1)-[r4:Capability_Capability]->(cap1)
with mainCap, adoptsCap1, cap1,
CASE WHEN capKPI IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: capKPI.ExternalID,
type: labels(capKPI)[0],
name: capKPI.Name,
description: capKPI.Description,
sectorID: capKPI.SectorID,
CCO: capKPI.CCO,
CFO: capKPI.CFO,
SVPHR: capKPI.SVPHR,
area: capKPI.Area,
CEO: capKPI.CEO,
CIO: capKPI.CIO,
COO: capKPI.COO
}
)
ELSE
[]
END as cpbltyIsMsrdByKpiPttrn
OPTIONAL MATCH (cap1)<-[r5:Capability_Capability]-(isPart1)
OPTIONAL MATCH (cap1)-[r6:KPI_Capability]->(KPI1)
OPTIONAL MATCH (cap1)-[r7:Capability_Capability]->(cap2)
with mainCap, adoptsCap1, cap1,cpbltyIsMsrdByKpiPttrn,cap2,
CASE WHEN KPI1 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: KPI1.ExternalID,
type: labels(KPI1)[0],
name: KPI1.Name,
description: KPI1.Description,
sectorID: KPI1.SectorID,
CCO: KPI1.CCO,
CFO: KPI1.CFO,
SVPHR: KPI1.SVPHR,
area: KPI1.Area,
CEO: KPI1.CEO,
CIO: KPI1.CIO,
COO: KPI1.COO
}
)
ELSE
[]
END as cpbltyIsMsrdByKpiPttrn1,
CASE WHEN isPart1 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: isPart1.ExternalID,
type: labels(isPart1)[0]
}
)
ELSE
[]
END as isPartOfCapability1
OPTIONAL MATCH (cap2)<-[r8:Capability_Capability]-(isPart2)
OPTIONAL MATCH (cap2)-[r9:KPI_Capability]->(KPI2)
OPTIONAL MATCH (cap2)-[r10:Capability_Capability]->(cap3)
with mainCap, adoptsCap1, cap1,cpbltyIsMsrdByKpiPttrn,cap2,cpbltyIsMsrdByKpiPttrn1,isPartOfCapability1,cap3,
CASE WHEN KPI2 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: KPI2.ExternalID,
type: labels(KPI2)[0],
name: KPI2.Name,
description: KPI2.Description,
sectorID: KPI2.SectorID,
CCO: KPI2.CCO,
CFO: KPI2.CFO,
SVPHR: KPI2.SVPHR,
area: KPI2.Area,
CEO: KPI2.CEO,
CIO: KPI2.CIO,
COO: KPI2.COO
}
)
ELSE
[]
END as cpbltyIsMsrdByKpiPttrn2,
CASE WHEN isPart2 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: isPart2.ExternalID,
type: labels(isPart2)[0]
}
)
ELSE
[]
END as isPartOfCapability2
OPTIONAL MATCH (cap3)<-[r11:Capability_Capability]-(isPart3)
OPTIONAL MATCH (cap3)-[r12:KPI_Capability]->(KPI3)
with mainCap, adoptsCap1, cap1,cpbltyIsMsrdByKpiPttrn,cap2,cpbltyIsMsrdByKpiPttrn1,isPartOfCapability1,cpbltyIsMsrdByKpiPttrn2,isPartOfCapability2,cap3,
CASE WHEN KPI3 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: KPI3.ExternalID,
type: labels(KPI3)[0],
name: KPI3.Name,
description: KPI3.Description,
sectorID: KPI3.SectorID,
CCO: KPI3.CCO,
CFO: KPI3.CFO,
SVPHR: KPI3.SVPHR,
area: KPI3.Area,
CEO: KPI3.CEO,
CIO: KPI3.CIO,
COO: KPI3.COO
}
)
ELSE
[]
END as cpbltyIsMsrdByKpiPttrn3,
CASE WHEN isPart3 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: isPart3.ExternalID,
type: labels(isPart3)[0]
}
)
ELSE
[]
END as isPartOfCapability3
with mainCap, adoptsCap1, cap1,cpbltyIsMsrdByKpiPttrn,cap2,cpbltyIsMsrdByKpiPttrn1,isPartOfCapability1,cpbltyIsMsrdByKpiPttrn2,isPartOfCapability2,
CASE WHEN cap3 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: cap3.ExternalID,
type: labels(cap3)[0],
name: cap3.Name,
description: cap3.Overview,
sectorID: cap3.SectorID,
cpbltyIsMsrdByKpiPttrn:cpbltyIsMsrdByKpiPttrn3,
isPartOfCapability:isPartOfCapability3
}
)
ELSE
[]
END as consistsOfCapability3
with mainCap, adoptsCap1, cap1,cpbltyIsMsrdByKpiPttrn,cpbltyIsMsrdByKpiPttrn1,isPartOfCapability1,
CASE WHEN cap2 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: cap2.ExternalID,
type: labels(cap2)[0],
name: cap2.Name,
description: cap2.Overview,
sectorID: cap2.SectorID,
cpbltyIsMsrdByKpiPttrn:cpbltyIsMsrdByKpiPttrn2,
isPartOfCapability:isPartOfCapability2,
consistsOfCapability:consistsOfCapability3
}
)
ELSE
[]
END as consistsOfCapability2
with mainCap, adoptsCap1, cpbltyIsMsrdByKpiPttrn,
CASE WHEN cap1 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: cap1.ExternalID,
type: labels(cap1)[0],
name: cap1.Name,
description: cap1.Overview,
sectorID: cap1.SectorID,
cpbltyIsMsrdByKpiPttrn:cpbltyIsMsrdByKpiPttrn1,
isPartOfCapability:isPartOfCapability1,
consistsOfCapability:consistsOfCapability2
}
)
ELSE
[]
END as consistsOfCapability1
OPTIONAL MATCH (mainCap)<-[r3:Capability_Capability]-(isPart:Capability)
return {
model:{
_id:mainCap.ExternalID,
type:labels(mainCap)[0],
name:mainCap.Name,
adoptsCapability:
CASE WHEN adoptsCap1 IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: adoptsCap1.ExternalID,
type: labels(adoptsCap1)[0],
name: adoptsCap1.Name,
description: adoptsCap1.Overview,
sectorID: adoptsCap1.SectorID,
cpbltyIsMsrdByKpiPttrn:cpbltyIsMsrdByKpiPttrn,
consistsOfCapability:consistsOfCapability1
}
)
ELSE
[]
END,
isPartOfCapability:
CASE WHEN isPart IS NOT NULL THEN
COLLECT(DISTINCT
{
_id: isPart.ExternalID,
type: labels(isPart)[0]
}
)
ELSE
[]
END
}
}
2 个答案:
答案 0 :(得分：1)
添加约束或索引
用于您的起始节点：
CREATE CONSTRAINT ON (mainCap:Capability) assert mainCap.SectorID IS UNIQUE;
或
CREATE INDEX ON :Capability(SectorID);
减少基数
最重要的一点是让你的中间基数降低。
否则，前一个语句产生的每一行都会重复匹配，无论它是否是重复的行。
MATCH (mainCap:Capability{SectorID:"PNU"})
OPTIONAL MATCH (mainCap)-[r1:CapabilityAdopts_Capability]->(adoptsCap1)-[r2:KPI_Capability]->(capKPI),
OPTIONAL MATCH (adoptsCap1)-[r4:Capability_Capability]->(cap1)
// reduce cardinality
with mainCap, adoptsCap1, cap1, collect(distinct capKPI) as _capKPI
OPTIONAL MATCH (cap1)<-[r5:Capability_Capability]-(isPart1)
with mainCap, adoptsCap1, cap1, _capKPI, collect(isPart1) as _isPart1
...
PS如果名称没有像mainCap as mainCap,
并尝试从顶部开始，随时进行剖析。
答案 1 :(得分：0)
根据您使用的Neo4j版本（在2.2.0之前或之后），尝试快速尝试：
尝试合并所有OPTIONAL MATCH条款，如下所示：
MATCH (mainCap:Capability{SectorID:"PNU"})
OPTIONAL MATCH
(mainCap)-[r1:CapabilityAdopts_Capability]->(adoptsCap1),
(adoptsCap1)-[r2:KPI_Capability]->(capKPI),
(adoptsCap1)-[r4:Capability_Capability]->(cap1),
(cap1)<-[r5:Capability_Capability]-(isPart1),
(cap1)-[r6:KPI_Capability]->(KPI1),
(cap1)-[r7:Capability_Capability]->(cap2),
(cap2)<-[r8:Capability_Capability]-(isPart2),
(cap2)-[r9:KPI_Capability]->(KPI2),
(cap2)-[r10:Capability_Capability]->(cap3),
(cap3)<-[r11:Capability_Capability]-(isPart3),
(cap3)-[r12:KPI_Capability]->(KPI3),
(mainCap)<-[r3:Capability_Capability]-(isPart:Capability)
而不是：
MATCH (mainCap:Capability{SectorID:"PNU"})
OPTIONAL MATCH (mainCap)-[r1:CapabilityAdopts_Capability]->(adoptsCap1)
OPTIONAL MATCH (adoptsCap1)-[r2:KPI_Capability]->(capKPI)
OPTIONAL MATCH (adoptsCap1)-[r4:Capability_Capability]->(cap1)
OPTIONAL MATCH (cap1)<-[r5:Capability_Capability]-(isPart1)
OPTIONAL MATCH (cap1)-[r6:KPI_Capability]->(KPI1)
OPTIONAL MATCH (cap1)-[r7:Capability_Capability]->(cap2)
OPTIONAL MATCH (cap2)<-[r8:Capability_Capability]-(isPart2)
OPTIONAL MATCH (cap2)-[r9:KPI_Capability]->(KPI2)
OPTIONAL MATCH (cap2)-[r10:Capability_Capability]->(cap3)
OPTIONAL MATCH (cap3)<-[r11:Capability_Capability]-(isPart3)
OPTIONAL MATCH (cap3)-[r12:KPI_Capability]->(KPI3)
OPTIONAL MATCH (mainCap)<-[r3:Capability_Capability]-(isPart:Capability)
修改
我没有仔细阅读查询，因为我打算打电话，但你也可以将它们组合起来：
OPTIONAL MATCH (mainCap)-[r1:CapabilityAdopts_Capability]->(adoptsCap1)-[r2:KPI_Capability]->(capKPI)
Neo4j非常聪明，无论如何都能很好地处理它，所以它是一种风格选择。
相关问题
最新问题
- 我写了这段代码，但我无法理解我的错误
- 我无法从一个代码实例的列表中删除 None 值，但我可以在另一个实例中。为什么它适用于一个细分市场而不适用于另一个细分市场？
- 是否有可能使 loadstring 不可能等于打印？卢阿
- java中的random.expovariate()
- Appscript 通过会议在 Google 日历中发送电子邮件和创建活动
- 为什么我的 Onclick 箭头功能在 React 中不起作用？
- 在此代码中是否有使用“this”的替代方法？
- 在 SQL Server 和 PostgreSQL 上查询，我如何从第一个表获得第二个表的可视化
- 每千个数字得到
- 更新了城市边界 KML 文件的来源？