Skip to content

Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure

TypeScript

test(desktop): distinguish csv totals from numbered columns

Alfredo Lopez8f377ef49c6c
diff --git a/desktop-app/src/__tests__/gfsLargeFileCsvFixture.test.ts b/desktop-app/src/__tests__/gfsLargeFileCsvFixture.test.ts
index 19947b87e..20bd6a050 100644
--- a/desktop-app/src/__tests__/gfsLargeFileCsvFixture.test.ts
+++ b/desktop-app/src/__tests__/gfsLargeFileCsvFixture.test.ts
@@ -114,6 +114,15 @@ describe('large-file CSV summary assertions', () => {
['Columnas: 78,59', []],
['Columnas: -78', []],
['Columnas: 78; total: 79 columnas', [78, 79]],
+ ['Columnas: 1. field_0, 2. field_1', []],
+ ['Columnas: 1) field_0, 2) field_1', []],
+ ['Columnas:1.field_0,2.field_1', []],
+ ['Hay 78 columnas. Columnas: 1) field_0, 2) field_1', [78]],
+ ['Columnas totales:79', [79]],
+ ['Columnas(total):79', [79]],
+ ['Columnas (total): 79', [79]],
+ ['Columnas:79. Nombres: field_0, field_1', [79]],
+ ['Columnas (78): 1. field_0, 2. field_1', [78]],
] as const)('extracts explicit column-count claims from %j', (summary, claims) => {
expect(csvColumnCountClaims(summary)).toEqual(claims)
})
diff --git a/desktop-app/test/e2e-playwright/helpers/gfsLargeFileCsvFixture.ts b/desktop-app/test/e2e-playwright/helpers/gfsLargeFileCsvFixture.ts
index 26282efca..6bca4d856 100644
--- a/desktop-app/test/e2e-playwright/helpers/gfsLargeFileCsvFixture.ts
+++ b/desktop-app/test/e2e-playwright/helpers/gfsLargeFileCsvFixture.ts
@@ -49,16 +49,25 @@ export function hasCsvDataRecordCount(summary: string, count: number): boolean {
export function csvColumnCountClaims(summary: string): number[] {
const label = String.raw`(?:columnas|columns)`
const number = String.raw`(?<![\p{L}\p{N}_\-\u2212])(\d{1,3}(?:[., \t]\d{3})+|\d+)(?![\p{L}\p{N}_]|[.,]\d)`
+ const total = String.raw`(?:total(?:es)?|\([ \t]*total[ \t]*\))`
// Horizontal spacing cannot mistake the first numbered item below a
// "Columns:" heading for a count of columns.
const pattern = new RegExp(
- String.raw`(?:\b${label}\b\*{0,2}[ \t]*(?:\(|:|es|=|de|son)?[ \t]*\*{0,2}[ \t]*${number}|${number}[ \t*\x60_]{1,12}\b${label}\b)`,
+ String.raw`(?:\b${label}\b\*{0,2}[ \t]*(?:${total}[ \t]*)?(?:\(|:|es|=|de|son)?[ \t]*\*{0,2}[ \t]*${number}|${number}[ \t*\x60_]{1,12}\b${label}\b)`,
'giu'
)
- return [...summary.matchAll(pattern)]
- .flatMap(match => [match[1], match[2]])
- .filter((raw): raw is string => Boolean(raw))
- .map(raw => Number(raw.replace(/[., \t]/g, '')))
+ const claims: number[] = []
+ for (const match of summary.matchAll(pattern)) {
+ const raw = match[1] ?? match[2]
+ if (!raw) continue
+ const explicit = /\btotal(?:es)?\b/iu.test(match[0]) || match[0].includes('(')
+ const tail = summary.slice(match.index + match[0].length)
+ const ordinal = /^[*_\x60]*[.)][ \t]*\S/u.test(tail)
+ const nextLabel = /^[*_\x60]*\.[ \t]*[*_\x60]*\p{L}[\p{L}\p{N}_ \t-]*[*_\x60]*:/u.test(tail)
+ if (match[1] && !explicit && ordinal && !nextLabel) continue
+ claims.push(Number(raw.replace(/[., \t]/g, '')))
+ }
+ return claims
}
/** Count distinct header mentions without returning private field names. */