web/ui/module/codemirror-promql/src/complete/hybrid.ts

// Copyright 2021 The Prometheus Authors
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

import { CompleteStrategy } from './index';
import { SyntaxNode } from '@lezer/common';
import { PrometheusClient } from '../client';
import {
  Add,
  AggregateExpr,
  And,
  BinaryExpr,
  BoolModifier,
  Div,
  Duration,
  Eql,
  EqlRegex,
  EqlSingle,
  FunctionCallBody,
  GroupingLabels,
  Gte,
  Gtr,
  LabelMatcher,
  LabelMatchers,
  LabelName,
  Lss,
  Lte,
  MatchOp,
  MatrixSelector,
  Identifier,
  Mod,
  Mul,
  Neq,
  NeqRegex,
  NumberLiteral,
  OffsetExpr,
  Or,
  Pow,
  PromQL,
  StepInvariantExpr,
  StringLiteral,
  Sub,
  SubqueryExpr,
  Unless,
  VectorSelector,
} from '@prometheus-io/lezer-promql';
import { Completion, CompletionContext, CompletionResult } from '@codemirror/autocomplete';
import { EditorState } from '@codemirror/state';
import { buildLabelMatchers, containsAtLeastOneChild, containsChild, walkBackward } from '../parser';
import {
  aggregateOpModifierTerms,
  aggregateOpTerms,
  atModifierTerms,
  binOpModifierTerms,
  binOpTerms,
  durationTerms,
  functionIdentifierTerms,
  matchOpTerms,
  numberTerms,
  snippets,
} from './promql.terms';
import { Matcher } from '../types';
import { syntaxTree } from '@codemirror/language';

const autocompleteNodes: { [key: string]: Completion[] } = {
  matchOp: matchOpTerms,
  binOp: binOpTerms,
  duration: durationTerms,
  binOpModifier: binOpModifierTerms,
  atModifier: atModifierTerms,
  functionIdentifier: functionIdentifierTerms,
  aggregateOp: aggregateOpTerms,
  aggregateOpModifier: aggregateOpModifierTerms,
  number: numberTerms,
};

// ContextKind is the different possible value determinate by the autocompletion
export enum ContextKind {
  // dynamic autocompletion (required a distant server)
  MetricName,
  LabelName,
  LabelValue,
  // static autocompletion
  Function,
  Aggregation,
  BinOpModifier,
  BinOp,
  MatchOp,
  AggregateOpModifier,
  Duration,
  Offset,
  Bool,
  AtModifiers,
  Number,
}

export interface Context {
  kind: ContextKind;
  metricName?: string;
  labelName?: string;
  matchers?: Matcher[];
}

function getMetricNameInGroupBy(tree: SyntaxNode, state: EditorState): string {
  // There should be an AggregateExpr as parent of the GroupingLabels.
  // Then we should find the VectorSelector child to be able to find the metric name.
  const currentNode: SyntaxNode | null = walkBackward(tree, AggregateExpr);
  if (!currentNode) {
    return '';
  }
  let metricName = '';
  currentNode.cursor().iterate((node) => {
    // Continue until we find the VectorSelector, then look up the metric name.
    if (node.type.id === VectorSelector) {
      metricName = getMetricNameInVectorSelector(node.node, state);
      if (metricName) {
        return false;
      }
    }
  });
  return metricName;
}

function getMetricNameInVectorSelector(tree: SyntaxNode, state: EditorState): string {
  // Find if there is a defined metric name. Should be used to autocomplete a labelValue or a labelName
  // First find the parent "VectorSelector" to be able to find then the subChild "Identifier" if it exists.
  let currentNode: SyntaxNode | null = walkBackward(tree, VectorSelector);
  if (!currentNode) {
    // Weird case that shouldn't happen, because "VectorSelector" is by definition the parent of the LabelMatchers.
    return '';
  }
  currentNode = currentNode.getChild(Identifier);
  if (!currentNode) {
    return '';
  }
  return state.sliceDoc(currentNode.from, currentNode.to);
}

function arrayToCompletionResult(data: Completion[], from: number, to: number, includeSnippet = false, span = true): CompletionResult {
  const options = data;
  if (includeSnippet) {
    options.push(...snippets);
  }
  return {
    from: from,
    to: to,
    options: options,
    validFor: span ? /^[a-zA-Z0-9_:]+$/ : undefined,
  } as CompletionResult;
}

// computeStartCompleteLabelPositionInLabelMatcherOrInGroupingLabel calculates the start position only when the node is a LabelMatchers or a GroupingLabels
function computeStartCompleteLabelPositionInLabelMatcherOrInGroupingLabel(node: SyntaxNode, pos: number): number {
  // Here we can have two different situations:
  // 1. `metric{}` or `sum by()` with the cursor between the bracket
  // and so we have increment the starting position to avoid to consider the open bracket when filtering the autocompletion list.
  // 2. `metric{foo="bar",} or `sum by(foo,)  with the cursor after the comma.
  // Then the start number should be the current position to avoid to consider the previous labelMatcher/groupingLabel when filtering the autocompletion list.
  let start = node.from + 1;
  if (node.firstChild !== null) {
    // here that means the LabelMatchers / GroupingLabels has a child, which is not possible if we have the expression `metric{}`. So we are likely trying to autocomplete the label list after a comma
    start = pos;
  }
  return start;
}

// computeStartCompletePosition calculates the start position of the autocompletion.
// It is an important step because the start position will be used by CMN to find the string and then to use it to filter the CompletionResult.
// A wrong `start` position will lead to have the completion not working.
// Note: this method is exported only for testing purpose.
export function computeStartCompletePosition(node: SyntaxNode, pos: number): number {
  let start = node.from;
  if (node.type.id === LabelMatchers || node.type.id === GroupingLabels) {
    start = computeStartCompleteLabelPositionInLabelMatcherOrInGroupingLabel(node, pos);
  } else if (node.type.id === FunctionCallBody || (node.type.id === StringLiteral && node.parent?.type.id === LabelMatcher)) {
    // When the cursor is between bracket, quote, we need to increment the starting position to avoid to consider the open bracket/ first string.
    start++;
  } else if (
    node.type.id === OffsetExpr ||
    (node.type.id === NumberLiteral && node.parent?.type.id === 0 && node.parent.parent?.type.id === SubqueryExpr) ||
    (node.type.id === 0 &&
      (node.parent?.type.id === OffsetExpr ||
        node.parent?.type.id === MatrixSelector ||
        (node.parent?.type.id === SubqueryExpr && containsAtLeastOneChild(node.parent, Duration))))
  ) {
    start = pos;
  }
  return start;
}

// analyzeCompletion is going to determinate what should be autocompleted.
// The value of the autocompletion is then calculate by the function buildCompletion.
// Note: this method is exported for testing purpose only. Do not use it directly.
export function analyzeCompletion(state: EditorState, node: SyntaxNode): Context[] {
  const result: Context[] = [];
  switch (node.type.id) {
    case 0: // 0 is the id of the error node
      if (node.parent?.type.id === OffsetExpr) {
        // we are likely in the given situation:
        // `metric_name offset 5` that leads to this tree:
        // `OffsetExpr(VectorSelector(Identifier),Offset,⚠)`
        // Here we can just autocomplete a duration.
        result.push({ kind: ContextKind.Duration });
        break;
      }
      if (node.parent?.type.id === LabelMatcher) {
        // In this case the current token is not itself a valid match op yet:
        //      metric_name{labelName!}
        result.push({ kind: ContextKind.MatchOp });
        break;
      }
      if (node.parent?.type.id === MatrixSelector) {
        // we are likely in the given situation:
        // `metric_name{}[5]`
        // We can also just autocomplete a duration
        result.push({ kind: ContextKind.Duration });
        break;
      }
      if (node.parent?.type.id === SubqueryExpr && containsAtLeastOneChild(node.parent, Duration)) {
        // we are likely in the given situation:
        //    `rate(foo[5d:5])`
        // so we should autocomplete a duration
        result.push({ kind: ContextKind.Duration });
        break;
      }
      // when we are in the situation 'metric_name !', we have the following tree
      // VectorSelector(Identifier,⚠)
      // We should try to know if the char '!' is part of a binOp.
      // Note: as it is quite experimental, maybe it requires more condition and to check the current tree (parent, other child at the same level ..etc.).
      const operator = state.sliceDoc(node.from, node.to);
      if (binOpTerms.filter((term) => term.label.includes(operator)).length > 0) {
        result.push({ kind: ContextKind.BinOp });
      }
      break;
    case Identifier:
      // sometimes an Identifier has an error has parent. This should be treated in priority
      if (node.parent?.type.id === 0) {
        const errorNodeParent = node.parent.parent;
        if (errorNodeParent?.type.id === StepInvariantExpr) {
          // we are likely in the given situation:
          //   `expr @ s`
          // we can autocomplete start / end
          result.push({ kind: ContextKind.AtModifiers });
          break;
        }
        if (errorNodeParent?.type.id === AggregateExpr) {
          // it matches 'sum() b'. So here we can autocomplete:
          // - the aggregate operation modifier
          // - the binary operation (since it's not mandatory to have an aggregate operation modifier)
          result.push({ kind: ContextKind.AggregateOpModifier }, { kind: ContextKind.BinOp });
          break;
        }
        if (errorNodeParent?.type.id === VectorSelector) {
          // it matches 'sum b'. So here we also have to autocomplete the aggregate operation modifier only
          // if the associated identifier is matching an aggregation operation.
          // Note: here is the corresponding tree in order to understand the situation:
          // VectorSelector(
          //   Identifier,
          //   ⚠(Identifier)
          // )
          const operator = getMetricNameInVectorSelector(node, state);
          if (aggregateOpTerms.filter((term) => term.label === operator).length > 0) {
            result.push({ kind: ContextKind.AggregateOpModifier });
          }
          // It's possible it also match the expr 'metric_name unle'.
          // It's also possible that the operator is also a metric even if it matches the list of aggregation function.
          // So we also have to autocomplete the binary operator.
          //
          // The expr `metric_name off` leads to the same tree. So we have to provide the offset keyword too here.
          result.push({ kind: ContextKind.BinOp }, { kind: ContextKind.Offset });
          break;
        }

        if (errorNodeParent && containsChild(errorNodeParent, 'Expr')) {
          // this last case can appear with the following expression:
          // 1. http_requests_total{method="GET"} off
          // 2. rate(foo[5m]) un
          // 3. sum(http_requests_total{method="GET"} off)
          // For these different cases we have this kind of tree:
          // Parent (
          //    ⚠(Identifier)
          // )
          // We don't really care about the parent, here we are more interested if in the siblings of the error node, there is the node 'Expr'
          // If it is the case, then likely we should autocomplete the BinOp or the offset.
          result.push({ kind: ContextKind.BinOp }, { kind: ContextKind.Offset });
          break;
        }
      }
      // As the leaf Identifier is coming for different cases, we have to take a bit time to analyze the tree
      // in order to know what we have to autocomplete exactly.
      // Here is some cases:
      // 1. metric_name / ignor --> we should autocomplete the BinOpModifier + metric/function/aggregation
      // 2. sum(http_requests_total{method="GET"} / o) --> BinOpModifier + metric/function/aggregation
      // Examples above give a different tree each time and ends up to be treated in this case.
      // But they all have the following common tree pattern:
      // Parent( ...,
      //         ... ,
      //         VectorSelector(Identifier)
      //       )
      //
      // So the first things to do is to get the `Parent` and to determinate if we are in this configuration.
      // Otherwise we would just have to autocomplete the metric / function / aggregation.

      const parent = node.parent?.parent;
      if (!parent) {
        // this case can be possible if the topNode is not anymore PromQL but MetricName.
        // In this particular case, then we just want to autocomplete the metric
        result.push({ kind: ContextKind.MetricName, metricName: state.sliceDoc(node.from, node.to) });
        break;
      }
      // now we have to know if we have two Expr in the direct children of the `parent`
      const containExprTwice = containsChild(parent, 'Expr', 'Expr');
      if (containExprTwice) {
        if (parent.type.id === BinaryExpr && !containsAtLeastOneChild(parent, 0)) {
          // We are likely in the case 1 or 5
          result.push(
            { kind: ContextKind.MetricName, metricName: state.sliceDoc(node.from, node.to) },
            { kind: ContextKind.Function },
            { kind: ContextKind.Aggregation },
            { kind: ContextKind.BinOpModifier },
            { kind: ContextKind.Number }
          );
          // in  case the BinaryExpr is a comparison, we should autocomplete the `bool` keyword. But only if it is not present.
          // When the `bool` keyword is NOT present, then the expression looks like this:
          // 			BinaryExpr( ..., Gtr , ... )
          // When the `bool` keyword is present, then the expression looks like this:
          //      BinaryExpr( ..., Gtr , BoolModifier(...), ... )
          if (containsAtLeastOneChild(parent, Eql, Gte, Gtr, Lte, Lss, Neq) && !containsAtLeastOneChild(parent, BoolModifier)) {
            result.push({ kind: ContextKind.Bool });
          }
        }
      } else {
        result.push(
          { kind: ContextKind.MetricName, metricName: state.sliceDoc(node.from, node.to) },
          { kind: ContextKind.Function },
          { kind: ContextKind.Aggregation }
        );
        if (parent.type.id !== FunctionCallBody && parent.type.id !== MatrixSelector) {
          // it's too avoid to autocomplete a number in situation where it shouldn't.
          // Like with `sum by(rat)`
          result.push({ kind: ContextKind.Number });
        }
      }
      break;
    case PromQL:
      if (node.firstChild !== null && node.firstChild.type.id === 0) {
        // this situation can happen when there is nothing in the text area and the user is explicitly triggering the autocompletion (with ctrl + space)
        result.push(
          { kind: ContextKind.MetricName, metricName: '' },
          { kind: ContextKind.Function },
          { kind: ContextKind.Aggregation },
          { kind: ContextKind.Number }
        );
      }
      break;
    case GroupingLabels:
      // In this case we are in the given situation:
      //      sum by () or sum (metric_name) by ()
      // so we have or to autocomplete any kind of labelName or to autocomplete only the labelName associated to the metric
      result.push({ kind: ContextKind.LabelName, metricName: getMetricNameInGroupBy(node, state) });
      break;
    case LabelMatchers:
      // In that case we are in the given situation:
      //       metric_name{} or {}
      // so we have or to autocomplete any kind of labelName or to autocomplete only the labelName associated to the metric
      result.push({ kind: ContextKind.LabelName, metricName: getMetricNameInVectorSelector(node, state) });
      break;
    case LabelName:
      if (node.parent?.type.id === GroupingLabels) {
        // In this case we are in the given situation:
        //      sum by (myL)
        // So we have to continue to autocomplete any kind of labelName
        result.push({ kind: ContextKind.LabelName });
      } else if (node.parent?.type.id === LabelMatcher) {
        // In that case we are in the given situation:
        //       metric_name{myL} or {myL}
        // so we have or to continue to autocomplete any kind of labelName or
        // to continue to autocomplete only the labelName associated to the metric
        result.push({ kind: ContextKind.LabelName, metricName: getMetricNameInVectorSelector(node, state) });
      }
      break;
    case StringLiteral:
      if (node.parent?.type.id === LabelMatcher) {
        // In this case we are in the given situation:
        //      metric_name{labelName=""}
        // So we can autocomplete the labelValue

        // Get the labelName.
        // By definition it's the firstChild: https://github.com/promlabs/lezer-promql/blob/0ef65e196a8db6a989ff3877d57fd0447d70e971/src/promql.grammar#L250
        let labelName = '';
        if (node.parent.firstChild?.type.id === LabelName) {
          labelName = state.sliceDoc(node.parent.firstChild.from, node.parent.firstChild.to);
        }
        // then find the metricName if it exists
        const metricName = getMetricNameInVectorSelector(node, state);
        // finally get the full matcher available
        const matcherNode = walkBackward(node, LabelMatchers);
        const labelMatchers = buildLabelMatchers(matcherNode ? matcherNode.getChildren(LabelMatcher) : [], state);
        result.push({
          kind: ContextKind.LabelValue,
          metricName: metricName,
          labelName: labelName,
          matchers: labelMatchers,
        });
      }
      break;
    case NumberLiteral:
      if (node.parent?.type.id === 0 && node.parent.parent?.type.id === SubqueryExpr) {
        // Here we are likely in this situation:
        //     `go[5d:4]`
        // and we have the given tree:
        // SubqueryExpr(
        //   VectorSelector(Identifier),
        //   Duration, Duration, ⚠(NumberLiteral)
        // )
        // So we should continue to autocomplete a duration
        result.push({ kind: ContextKind.Duration });
      } else {
        result.push({ kind: ContextKind.Number });
      }
      break;
    case Duration:
    case OffsetExpr:
      result.push({ kind: ContextKind.Duration });
      break;
    case FunctionCallBody:
      // In this case we are in the given situation:
      //       sum() or in rate()
      // with the cursor between the bracket. So we can autocomplete the metric, the function and the aggregation.
      result.push({ kind: ContextKind.MetricName, metricName: '' }, { kind: ContextKind.Function }, { kind: ContextKind.Aggregation });
      break;
    case Neq:
      if (node.parent?.type.id === MatchOp) {
        result.push({ kind: ContextKind.MatchOp });
      } else if (node.parent?.type.id === BinaryExpr) {
        result.push({ kind: ContextKind.BinOp });
      }
      break;
    case EqlSingle:
    case EqlRegex:
    case NeqRegex:
    case MatchOp:
      result.push({ kind: ContextKind.MatchOp });
      break;
    case Pow:
    case Mul:
    case Div:
    case Mod:
    case Add:
    case Sub:
    case Eql:
    case Gte:
    case Gtr:
    case Lte:
    case Lss:
    case And:
    case Unless:
    case Or:
    case BinaryExpr:
      result.push({ kind: ContextKind.BinOp });
      break;
  }
  return result;
}

// HybridComplete provides a full completion result with or without a remote prometheus.
export class HybridComplete implements CompleteStrategy {
  private readonly prometheusClient: PrometheusClient | undefined;
  private readonly maxMetricsMetadata: number;

  constructor(prometheusClient?: PrometheusClient, maxMetricsMetadata = 10000) {
    this.prometheusClient = prometheusClient;
    this.maxMetricsMetadata = maxMetricsMetadata;
  }

  getPrometheusClient(): PrometheusClient | undefined {
    return this.prometheusClient;
  }

  promQL(context: CompletionContext): Promise<CompletionResult | null> | CompletionResult | null {
    const { state, pos } = context;
    const tree = syntaxTree(state).resolve(pos, -1);
    const contexts = analyzeCompletion(state, tree);
    let asyncResult: Promise<Completion[]> = Promise.resolve([]);
    let completeSnippet = false;
    let span = true;
    for (const context of contexts) {
      switch (context.kind) {
        case ContextKind.Aggregation:
          completeSnippet = true;
          asyncResult = asyncResult.then((result) => {
            return result.concat(autocompleteNodes.aggregateOp);
          });
          break;
        case ContextKind.Function:
          completeSnippet = true;
          asyncResult = asyncResult.then((result) => {
            return result.concat(autocompleteNodes.functionIdentifier);
          });
          break;
        case ContextKind.BinOpModifier:
          asyncResult = asyncResult.then((result) => {
            return result.concat(autocompleteNodes.binOpModifier);
          });
          break;
        case ContextKind.BinOp:
          asyncResult = asyncResult.then((result) => {
            return result.concat(autocompleteNodes.binOp);
          });
          break;
        case ContextKind.MatchOp:
          asyncResult = asyncResult.then((result) => {
            return result.concat(autocompleteNodes.matchOp);
          });
          break;
        case ContextKind.AggregateOpModifier:
          asyncResult = asyncResult.then((result) => {
            return result.concat(autocompleteNodes.aggregateOpModifier);
          });
          break;
        case ContextKind.Duration:
          span = false;
          asyncResult = asyncResult.then((result) => {
            return result.concat(autocompleteNodes.duration);
          });
          break;
        case ContextKind.Offset:
          asyncResult = asyncResult.then((result) => {
            return result.concat([{ label: 'offset' }]);
          });
          break;
        case ContextKind.Bool:
          asyncResult = asyncResult.then((result) => {
            return result.concat([{ label: 'bool' }]);
          });
          break;
        case ContextKind.AtModifiers:
          asyncResult = asyncResult.then((result) => {
            return result.concat(autocompleteNodes.atModifier);
          });
          break;
        case ContextKind.Number:
          asyncResult = asyncResult.then((result) => {
            return result.concat(autocompleteNodes.number);
          });
          break;
        case ContextKind.MetricName:
          asyncResult = asyncResult.then((result) => {
            return this.autocompleteMetricName(result, context);
          });
          break;
        case ContextKind.LabelName:
          asyncResult = asyncResult.then((result) => {
            return this.autocompleteLabelName(result, context);
          });
          break;
        case ContextKind.LabelValue:
          asyncResult = asyncResult.then((result) => {
            return this.autocompleteLabelValue(result, context);
          });
      }
    }
    return asyncResult.then((result) => {
      return arrayToCompletionResult(result, computeStartCompletePosition(tree, pos), pos, completeSnippet, span);
    });
  }

  private autocompleteMetricName(result: Completion[], context: Context): Completion[] | Promise<Completion[]> {
    if (!this.prometheusClient) {
      return result;
    }
    const metricCompletion = new Map<string, Completion>();
    return this.prometheusClient
      .metricNames(context.metricName)
      .then((metricNames: string[]) => {
        for (const metricName of metricNames) {
          metricCompletion.set(metricName, { label: metricName, type: 'constant' });
        }

        // avoid to get all metric metadata if the prometheus server is too big
        if (metricNames.length <= this.maxMetricsMetadata) {
          // in order to enrich the completion list of the metric,
          // we are trying to find the associated metadata
          return this.prometheusClient?.metricMetadata();
        }
      })
      .then((metricMetadata) => {
        if (metricMetadata) {
          for (const [metricName, node] of metricCompletion) {
            // For histograms and summaries, the metadata is only exposed for the base metric name,
            // not separately for the _count, _sum, and _bucket time series.
            const metadata = metricMetadata[metricName.replace(/(_count|_sum|_bucket)$/, '')];
            if (metadata) {
              if (metadata.length > 1) {
                // it means the metricName has different possible helper and type
                for (const m of metadata) {
                  if (node.detail === '') {
                    node.detail = m.type;
                  } else if (node.detail !== m.type) {
                    node.detail = 'unknown';
                    node.info = 'multiple different definitions for this metric';
                  }

                  if (node.info === '') {
                    node.info = m.help;
                  } else if (node.info !== m.help) {
                    node.info = 'multiple different definitions for this metric';
                  }
                }
              } else if (metadata.length === 1) {
                let { type, help } = metadata[0];
                if (type === 'histogram' || type === 'summary') {
                  if (metricName.endsWith('_count')) {
                    type = 'counter';
                    help = `The total number of observations for: ${help}`;
                  }
                  if (metricName.endsWith('_sum')) {
                    type = 'counter';
                    help = `The total sum of observations for: ${help}`;
                  }
                  if (metricName.endsWith('_bucket')) {
                    type = 'counter';
                    help = `The total count of observations for a bucket in the histogram: ${help}`;
                  }
                }
                node.detail = type;
                node.info = help;
              }
            }
          }
        }
        return result.concat(Array.from(metricCompletion.values()));
      });
  }

  private autocompleteLabelName(result: Completion[], context: Context): Completion[] | Promise<Completion[]> {
    if (!this.prometheusClient) {
      return result;
    }
    return this.prometheusClient.labelNames(context.metricName).then((labelNames: string[]) => {
      return result.concat(labelNames.map((value) => ({ label: value, type: 'constant' })));
    });
  }

  private autocompleteLabelValue(result: Completion[], context: Context): Completion[] | Promise<Completion[]> {
    if (!this.prometheusClient || !context.labelName) {
      return result;
    }
    return this.prometheusClient.labelValues(context.labelName, context.metricName, context.matchers).then((labelValues: string[]) => {
      return result.concat(labelValues.map((value) => ({ label: value, type: 'text' })));
    });
  }
}