Improve performance of the new function by a factor of 3-4x using a memoization technique.
Hunter Trujillo
parent
f88c43cdbb
commit
b7c9243f3d
|
|
@ -206,19 +206,19 @@ describe('timeSeriesToDygraph', () => {
|
|||
labels: [
|
||||
'time',
|
||||
`m1.f1`,
|
||||
`m1.f1-1`,
|
||||
`m1.f1`,
|
||||
],
|
||||
timeSeries: [
|
||||
[new Date(1000), 1, null],
|
||||
[new Date(2000), 2, 3],
|
||||
[new Date(4000), 4, null],
|
||||
[new Date(4000), null, 4],
|
||||
],
|
||||
dygraphSeries: {
|
||||
// 'm1.f1': {
|
||||
// axis: 'y',
|
||||
// strokeWidth,
|
||||
// },
|
||||
'm1.f1': {
|
||||
axis: 'y',
|
||||
strokeWidth,
|
||||
},
|
||||
'm1.f1-1': {
|
||||
axis: 'y2',
|
||||
strokeWidth,
|
||||
},
|
||||
|
|
|
|||
|
|
@ -1,181 +1,126 @@
|
|||
import _ from 'lodash';
|
||||
import {STROKE_WIDTH} from 'src/shared/constants';
|
||||
/**
|
||||
* Accepts an array of raw influxdb responses and returns a format
|
||||
* that Dygraph understands.
|
||||
*/
|
||||
|
||||
// activeQueryIndex is an optional argument that indicated which query's series
|
||||
// we want highlighted.
|
||||
// activeQueryIndex is an optional argument that indicated which query's series we want highlighted.
|
||||
export default function timeSeriesToDygraph(raw = [], activeQueryIndex, isInDataExplorer) {
|
||||
const labels = []; // all of the effective field names (i.e. <measurement>.<field>)
|
||||
const fieldToIndex = {}; // see parseSeries
|
||||
const dates = {}; // map of date as string to date value to minimize string coercion
|
||||
const dygraphSeries = {}; // dygraphSeries is a graph legend label and its corresponding y-axis e.g. {legendLabel1: 'y', legendLabel2: 'y2'};
|
||||
const d0 = new Date();
|
||||
|
||||
/**
|
||||
* dateToFieldValue will look like:
|
||||
*
|
||||
* {
|
||||
* Date1: {
|
||||
* effectiveFieldName_1: ValueForField1AtDate1,
|
||||
* effectiveFieldName_2: ValueForField2AtDate1,
|
||||
* ...
|
||||
* },
|
||||
* Date2: {
|
||||
* effectiveFieldName_1: ValueForField1AtDate2,
|
||||
* effectiveFieldName_2: ValueForField2AtDate2,
|
||||
* ...
|
||||
* }
|
||||
* }
|
||||
*/
|
||||
const dateToFieldValue = {};
|
||||
// collect results from each influx response
|
||||
const results = raw.reduce((acc, rawResponse, responseIndex) => {
|
||||
const responses = _.get(rawResponse, 'response.results', []);
|
||||
const indexedResponses = responses.map((response) => ({...response, responseIndex}));
|
||||
return [...acc, ...indexedResponses];
|
||||
}, []);
|
||||
|
||||
raw.forEach(({response}, queryIndex) => {
|
||||
// If a response is an empty result set or a query returned an error
|
||||
// from InfluxDB, don't try and parse.
|
||||
if (response.results.length) {
|
||||
if (isEmpty(response) || hasError(response)) {
|
||||
return;
|
||||
}
|
||||
}
|
||||
// collect each series
|
||||
const serieses = results.reduce((acc, {series = [], responseIndex}, index) => {
|
||||
return [...acc, ...series.map((item) => ({...item, responseIndex, index}))];
|
||||
}, []);
|
||||
|
||||
/**
|
||||
* response looks like:
|
||||
* {
|
||||
* results: [
|
||||
* { series: [...] },
|
||||
* { series: [...] },
|
||||
* ]
|
||||
* }
|
||||
*/
|
||||
response.results.forEach(parseResult);
|
||||
// convert series into cells with rows and columns
|
||||
const cells = serieses.reduce((acc, {name, columns, values, index, responseIndex, tags = {}}) => {
|
||||
const rows = values.map((vals) => ({
|
||||
name,
|
||||
vals,
|
||||
index,
|
||||
}));
|
||||
|
||||
function parseResult(s) {
|
||||
/*
|
||||
* s looks like:
|
||||
* {
|
||||
* series: [
|
||||
* {
|
||||
* name: "<measurement>",
|
||||
* columns: ["time", "<field name 1>", "<field name 2>", ...],
|
||||
* values: [<time>, <value of field 1>, <value of field 2>, ...],
|
||||
* },
|
||||
* }
|
||||
*/
|
||||
s.series.forEach(parseSeries);
|
||||
}
|
||||
columns.shift();
|
||||
|
||||
function parseSeries(series) {
|
||||
/*
|
||||
* series looks like:
|
||||
* {
|
||||
* name: "<measurement>",
|
||||
* columns: ["time", "<field name 1>", "<field name 2>", ...],
|
||||
* values: [
|
||||
* [<time1>, <value of field 1 @ time1>, <value of field 2 @ time1>, ...],
|
||||
* [<time2>, <value of field 1 @ time2>, <value of field 2 @ time2>, ...],
|
||||
* ]
|
||||
* }
|
||||
*/
|
||||
const measurementName = series.name;
|
||||
const columns = series.columns;
|
||||
// tagSet is each tag key and value for a series
|
||||
const tagSet = Object.keys(tags).map((tag) => `[${tag}=${tags[tag]}]`).sort().join('');
|
||||
|
||||
// Tags are only included in an influxdb response under certain circumstances, e.g.
|
||||
// when a query is using GROUP BY (<tag key>).
|
||||
const tags = Object.keys(series.tags || {}).map((key) => {
|
||||
return `[${key}=${series.tags[key]}]`;
|
||||
}).sort().join('');
|
||||
rows.forEach(({vals, name: measurement, index: seriesIndex}) => {
|
||||
const [time, ...rowValues] = vals;
|
||||
|
||||
columns.slice(1).forEach((fieldName) => {
|
||||
let effectiveFieldName = `${measurementName}.${fieldName}${tags}`;
|
||||
|
||||
// If there are duplicate effectiveFieldNames identify them by their queryIndex
|
||||
if (effectiveFieldName in dygraphSeries) {
|
||||
effectiveFieldName = `${effectiveFieldName}-${queryIndex}`;
|
||||
}
|
||||
|
||||
// Given a field name, identify which column in the timeSeries result should hold the field's value
|
||||
// ex given this timeSeries [Date, 10, 20, 30] field index at 2 would correspond to value 20
|
||||
fieldToIndex[effectiveFieldName] = labels.length + 1;
|
||||
labels.push(effectiveFieldName);
|
||||
|
||||
const {light, heavy} = STROKE_WIDTH;
|
||||
|
||||
const dygraphSeriesStyles = {
|
||||
strokeWidth: queryIndex === activeQueryIndex ? heavy : light,
|
||||
};
|
||||
|
||||
if (!isInDataExplorer) {
|
||||
dygraphSeriesStyles.axis = queryIndex === 0 ? 'y' : 'y2';
|
||||
}
|
||||
|
||||
dygraphSeries[effectiveFieldName] = dygraphSeriesStyles;
|
||||
});
|
||||
|
||||
(series.values || []).forEach(parseRow);
|
||||
|
||||
function parseRow(row) {
|
||||
/**
|
||||
* row looks like:
|
||||
* [<time1>, <value of field 1 @ time1>, <value of field 2 @ time1>, ...]
|
||||
*/
|
||||
const date = row[0];
|
||||
const dateString = date.toString();
|
||||
row.forEach((value, index) => {
|
||||
if (index === 0) {
|
||||
// index 0 in a row is always the timestamp
|
||||
if (!dateToFieldValue[dateString]) {
|
||||
dateToFieldValue[dateString] = {};
|
||||
dates[dateString] = date;
|
||||
}
|
||||
return;
|
||||
}
|
||||
|
||||
const fieldName = columns[index];
|
||||
let effectiveFieldName = `${measurementName}.${fieldName}${tags}`;
|
||||
|
||||
// If there are duplicate effectiveFieldNames identify them by their queryIndex
|
||||
if (effectiveFieldName in dateToFieldValue[dateString]) {
|
||||
effectiveFieldName = `${effectiveFieldName}-${queryIndex}`;
|
||||
}
|
||||
|
||||
dateToFieldValue[dateString][effectiveFieldName] = value;
|
||||
rowValues.forEach((value, i) => {
|
||||
const field = columns[i];
|
||||
acc.push({
|
||||
label: `${measurement}.${field}${tagSet}`,
|
||||
value,
|
||||
time,
|
||||
seriesIndex,
|
||||
responseIndex,
|
||||
});
|
||||
}
|
||||
});
|
||||
});
|
||||
|
||||
return acc;
|
||||
}, []);
|
||||
|
||||
// labels are a unique combination of measurement, fields, and tags that indicate a specific series on the graph legend
|
||||
const labels = cells.reduce((acc, {label, seriesIndex, responseIndex}) => {
|
||||
const existingLabel = acc.find(({
|
||||
label: findLabel,
|
||||
seriesIndex: findSeriesIndex,
|
||||
}) => findLabel === label && findSeriesIndex === seriesIndex);
|
||||
|
||||
if (!existingLabel) {
|
||||
acc.push({
|
||||
label,
|
||||
seriesIndex,
|
||||
responseIndex,
|
||||
});
|
||||
}
|
||||
});
|
||||
|
||||
function buildTimeSeries() {
|
||||
const allDates = Object.keys(dateToFieldValue);
|
||||
allDates.sort((a, b) => a - b);
|
||||
const rowLength = labels.length + 1;
|
||||
return allDates.map((date) => {
|
||||
const row = new Array(rowLength);
|
||||
return acc;
|
||||
}, []);
|
||||
|
||||
row.fill(null);
|
||||
row[0] = new Date(dates[date]);
|
||||
const sortedLabels = _.sortBy(labels, 'label');
|
||||
const memo = {};
|
||||
|
||||
const fieldsForRow = dateToFieldValue[date];
|
||||
Object.keys(fieldsForRow).forEach((effectiveFieldName) => {
|
||||
row[fieldToIndex[effectiveFieldName]] = fieldsForRow[effectiveFieldName];
|
||||
const ts0 = new Date();
|
||||
|
||||
const timeSeries = cells.reduce((acc, cell) => {
|
||||
let existingRowIndex = memo[cell.time];
|
||||
|
||||
if (existingRowIndex === undefined) {
|
||||
acc.push({
|
||||
time: cell.time,
|
||||
values: Array(sortedLabels.length).fill(null),
|
||||
});
|
||||
|
||||
return row;
|
||||
});
|
||||
}
|
||||
existingRowIndex = acc.length - 1;
|
||||
memo[cell.time] = existingRowIndex;
|
||||
}
|
||||
|
||||
const values = acc[existingRowIndex].values;
|
||||
const labelIndex = sortedLabels.findIndex(({label, seriesIndex}) => label === cell.label && cell.seriesIndex === seriesIndex);;
|
||||
values[labelIndex] = cell.value;
|
||||
acc[existingRowIndex].values = values;
|
||||
|
||||
return acc;
|
||||
}, []);
|
||||
const ts1 = new Date();
|
||||
|
||||
const sortedTimeSeries = _.sortBy(timeSeries, 'time');
|
||||
|
||||
const {light, heavy} = STROKE_WIDTH;
|
||||
|
||||
const dygraphSeries = sortedLabels.reduce((acc, {label, responseIndex}) => {
|
||||
acc[label] = {
|
||||
strokeWidth: responseIndex === activeQueryIndex ? heavy : light,
|
||||
};
|
||||
|
||||
if (!isInDataExplorer) {
|
||||
acc[label].axis = responseIndex === 0 ? 'y' : 'y2';
|
||||
}
|
||||
|
||||
return acc;
|
||||
}, {});
|
||||
|
||||
const d1 = new Date();
|
||||
|
||||
console.log(`new function took ${d1 - d0}ms to run for ${cells.length} cells.`);
|
||||
console.log(`timeSeries took ${ts1 - ts0}ms to run for ${timeSeries.length} records.`);
|
||||
|
||||
return {
|
||||
labels: ['time', ...labels.sort()],
|
||||
timeSeries: buildTimeSeries(),
|
||||
labels: ["time", ...sortedLabels.map(({label}) => label)],
|
||||
timeSeries: sortedTimeSeries.map(({time, values}) => ([new Date(time), ...values])),
|
||||
dygraphSeries,
|
||||
};
|
||||
}
|
||||
|
||||
function isEmpty(resp) {
|
||||
return !resp.results[0].series;
|
||||
}
|
||||
|
||||
function hasError(resp) {
|
||||
return !!resp.results[0].error;
|
||||
}
|
||||
|
||||
|
|
|
|||
Loading…
Reference in New Issue