core: read sessions with line reader instead of bufio.Scanner
Loading or exporting a session containing very large JSONL rows (image blocks, big tool outputs, compacted history) failed with 'bufio.Scanner: token too long' — Scanner caps each token to its buffer size, even when bumped to 20 MiB. A single oversized row blocked OpenSession entirely so an existing long session could not be resumed. Switch session readers (OpenSession, SessionUsage, describeSession, sessionHasNoMessages, ExportSession, ImportSession, BranchSession, firstUserPrompt) to a shared bufio.Reader.ReadBytes-based JSONL helper that handles arbitrarily long lines. Add a regression test that opens and exports a session containing a >20 MiB row.
- commit
- 1a49277130e730448b16d0284be4ee361fc9510a
- author
- patriceckhart · · 2026-05-04 10:14 UTC
- parent
- b45f107
- signature
- unsigned
- refs
- tag: v0.1.49
3 files · +188 -113
M internal/core/session.go +54 -35
| @@ -2,10 +2,12 @@ package core | |||
| 2 | 2 | ||
| 3 | 3 | import ( | |
| 4 | 4 | "bufio" | |
| 5 | + | "bytes" | |
| 5 | 6 | "crypto/sha256" | |
| 6 | 7 | "encoding/hex" | |
| 7 | 8 | "encoding/json" | |
| 8 | 9 | "fmt" | |
| 10 | + | "io" | |
| 9 | 11 | "os" | |
| 10 | 12 | "path/filepath" | |
| 11 | 13 | "sort" | |
| @@ -110,6 +112,27 @@ func NewSession(root, cwd, providerName, model, version string) (*Session, error | |||
| 110 | 112 | return s, nil | |
| 111 | 113 | } | |
| 112 | 114 | ||
| 115 | + | func forEachJSONLLine(r io.Reader, fn func([]byte) error) error { | |
| 116 | + | br := bufio.NewReader(r) | |
| 117 | + | for { | |
| 118 | + | line, err := br.ReadBytes('\n') | |
| 119 | + | if len(line) > 0 { | |
| 120 | + | line = bytes.TrimRight(line, "\r\n") | |
| 121 | + | if len(line) > 0 { | |
| 122 | + | if ferr := fn(line); ferr != nil { | |
| 123 | + | return ferr | |
| 124 | + | } | |
| 125 | + | } | |
| 126 | + | } | |
| 127 | + | if err == io.EOF { | |
| 128 | + | return nil | |
| 129 | + | } | |
| 130 | + | if err != nil { | |
| 131 | + | return err | |
| 132 | + | } | |
| 133 | + | } | |
| 134 | + | } | |
| 135 | + | ||
| 113 | 136 | // SessionUsage returns the most recent cumulative usage row stored in | |
| 114 | 137 | // a session file. Sessions append one usage row per completed turn; the | |
| 115 | 138 | // latest row's cumulative field is the session total. Missing usage rows | |
| @@ -122,21 +145,19 @@ func SessionUsage(path string) (provider.Usage, error) { | |||
| 122 | 145 | defer f.Close() | |
| 123 | 146 | ||
| 124 | 147 | var usage provider.Usage | |
| 125 | - | sc := bufio.NewScanner(f) | |
| 126 | - | sc.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | |
| 127 | - | for sc.Scan() { | |
| 148 | + | if err := forEachJSONLLine(f, func(line []byte) error { | |
| 128 | 149 | var head sessionLineHead | |
| 129 | - | if err := json.Unmarshal(sc.Bytes(), &head); err != nil || head.Type != "usage" { | |
| 130 | - | continue | |
| 150 | + | if err := json.Unmarshal(line, &head); err != nil || head.Type != "usage" { | |
| 151 | + | return nil | |
| 131 | 152 | } | |
| 132 | 153 | var row struct { | |
| 133 | 154 | Cumulative provider.Usage `json:"cumulative"` | |
| 134 | 155 | } | |
| 135 | - | if err := json.Unmarshal(sc.Bytes(), &row); err == nil { | |
| 156 | + | if err := json.Unmarshal(line, &row); err == nil { | |
| 136 | 157 | usage = row.Cumulative | |
| 137 | 158 | } | |
| 138 | - | } | |
| 139 | - | if err := sc.Err(); err != nil { | |
| 159 | + | return nil | |
| 160 | + | }); err != nil { | |
| 140 | 161 | return provider.Usage{}, err | |
| 141 | 162 | } | |
| 142 | 163 | return usage, nil | |
| @@ -152,28 +173,26 @@ func OpenSession(path string) (*Session, []provider.Message, error) { | |||
| 152 | 173 | ||
| 153 | 174 | var meta SessionMeta | |
| 154 | 175 | var messages []provider.Message | |
| 155 | - | sc := bufio.NewScanner(f) | |
| 156 | - | sc.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | |
| 157 | - | for sc.Scan() { | |
| 176 | + | if err := forEachJSONLLine(f, func(line []byte) error { | |
| 158 | 177 | var head sessionLineHead | |
| 159 | - | if err := json.Unmarshal(sc.Bytes(), &head); err != nil { | |
| 160 | - | continue | |
| 178 | + | if err := json.Unmarshal(line, &head); err != nil { | |
| 179 | + | return nil | |
| 161 | 180 | } | |
| 162 | 181 | switch head.Type { | |
| 163 | 182 | case "meta": | |
| 164 | 183 | var row struct { | |
| 165 | 184 | Meta SessionMeta `json:"meta"` | |
| 166 | 185 | } | |
| 167 | - | if err := json.Unmarshal(sc.Bytes(), &row); err == nil { | |
| 186 | + | if err := json.Unmarshal(line, &row); err == nil { | |
| 168 | 187 | meta = row.Meta | |
| 169 | 188 | } | |
| 170 | 189 | case "message": | |
| 171 | - | if msg, err := hydrateMessage(sc.Bytes()); err == nil && len(msg.Content) > 0 { | |
| 190 | + | if msg, err := hydrateMessage(line); err == nil && len(msg.Content) > 0 { | |
| 172 | 191 | messages = append(messages, msg) | |
| 173 | 192 | } | |
| 174 | 193 | } | |
| 175 | - | } | |
| 176 | - | if err := sc.Err(); err != nil { | |
| 194 | + | return nil | |
| 195 | + | }); err != nil { | |
| 177 | 196 | return nil, nil, err | |
| 178 | 197 | } | |
| 179 | 198 | messages = repairToolUseResultPairs(messages) | |
| @@ -333,19 +352,17 @@ func describeSession(path string) SessionSummary { | |||
| 333 | 352 | return s | |
| 334 | 353 | } | |
| 335 | 354 | defer f.Close() | |
| 336 | - | sc := bufio.NewScanner(f) | |
| 337 | - | sc.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | |
| 338 | - | for sc.Scan() { | |
| 355 | + | _ = forEachJSONLLine(f, func(line []byte) error { | |
| 339 | 356 | var head sessionLineHead | |
| 340 | - | if err := json.Unmarshal(sc.Bytes(), &head); err != nil { | |
| 341 | - | continue | |
| 357 | + | if err := json.Unmarshal(line, &head); err != nil { | |
| 358 | + | return nil | |
| 342 | 359 | } | |
| 343 | 360 | switch head.Type { | |
| 344 | 361 | case "meta": | |
| 345 | 362 | var row struct { | |
| 346 | 363 | Meta SessionMeta `json:"meta"` | |
| 347 | 364 | } | |
| 348 | - | if err := json.Unmarshal(sc.Bytes(), &row); err == nil { | |
| 365 | + | if err := json.Unmarshal(line, &row); err == nil { | |
| 349 | 366 | s.Started = row.Meta.Started | |
| 350 | 367 | s.Model = row.Meta.Model | |
| 351 | 368 | s.Provider = row.Meta.Provider | |
| @@ -354,24 +371,25 @@ func describeSession(path string) SessionSummary { | |||
| 354 | 371 | case "message": | |
| 355 | 372 | s.MessageCount++ | |
| 356 | 373 | if s.FirstUserText == "" { | |
| 357 | - | s.FirstUserText = firstUserText(sc.Bytes()) | |
| 374 | + | s.FirstUserText = firstUserText(line) | |
| 358 | 375 | } | |
| 359 | 376 | case "rename": | |
| 360 | 377 | var row struct { | |
| 361 | 378 | Title string `json:"title"` | |
| 362 | 379 | } | |
| 363 | - | if err := json.Unmarshal(sc.Bytes(), &row); err == nil && row.Title != "" { | |
| 380 | + | if err := json.Unmarshal(line, &row); err == nil && row.Title != "" { | |
| 364 | 381 | s.Title = row.Title | |
| 365 | 382 | } | |
| 366 | 383 | case "usage": | |
| 367 | 384 | var row struct { | |
| 368 | 385 | Cumulative provider.Usage `json:"cumulative"` | |
| 369 | 386 | } | |
| 370 | - | if err := json.Unmarshal(sc.Bytes(), &row); err == nil { | |
| 387 | + | if err := json.Unmarshal(line, &row); err == nil { | |
| 371 | 388 | s.TotalCost = row.Cumulative.CostUSD | |
| 372 | 389 | } | |
| 373 | 390 | } | |
| 374 | - | } | |
| 391 | + | return nil | |
| 392 | + | }) | |
| 375 | 393 | return s | |
| 376 | 394 | } | |
| 377 | 395 | ||
| @@ -429,18 +447,19 @@ func sessionHasNoMessages(path string) bool { | |||
| 429 | 447 | return false | |
| 430 | 448 | } | |
| 431 | 449 | defer f.Close() | |
| 432 | - | sc := bufio.NewScanner(f) | |
| 433 | - | sc.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | |
| 434 | - | for sc.Scan() { | |
| 450 | + | hasMessage := false | |
| 451 | + | _ = forEachJSONLLine(f, func(line []byte) error { | |
| 435 | 452 | var head sessionLineHead | |
| 436 | - | if err := json.Unmarshal(sc.Bytes(), &head); err != nil { | |
| 437 | - | continue | |
| 453 | + | if err := json.Unmarshal(line, &head); err != nil { | |
| 454 | + | return nil | |
| 438 | 455 | } | |
| 439 | 456 | if head.Type == "message" { | |
| 440 | - | return false | |
| 457 | + | hasMessage = true | |
| 458 | + | return io.EOF | |
| 441 | 459 | } | |
| 442 | - | } | |
| 443 | - | return true | |
| 460 | + | return nil | |
| 461 | + | }) | |
| 462 | + | return !hasMessage | |
| 444 | 463 | } | |
| 445 | 464 | ||
| 446 | 465 | // ListSessions returns session file paths for cwd, most-recently- | |
| @@ -2,10 +2,12 @@ package core | |||||
| 2 | 2 | ||||
| 3 | import ( | 3 | import ( | ||
| 4 | "bufio" | 4 | "bufio" | ||
| 5 | + | "bytes" | |||
| 5 | "crypto/sha256" | 6 | "crypto/sha256" | ||
| 6 | "encoding/hex" | 7 | "encoding/hex" | ||
| 7 | "encoding/json" | 8 | "encoding/json" | ||
| 8 | "fmt" | 9 | "fmt" | ||
| 10 | + | "io" | |||
| 9 | "os" | 11 | "os" | ||
| 10 | "path/filepath" | 12 | "path/filepath" | ||
| 11 | "sort" | 13 | "sort" | ||
| @@ -110,6 +112,27 @@ func NewSession(root, cwd, providerName, model, version string) (*Session, error | |||||
| 110 | return s, nil | 112 | return s, nil | ||
| 111 | } | 113 | } | ||
| 112 | 114 | ||||
| 115 | + | func forEachJSONLLine(r io.Reader, fn func([]byte) error) error { | |||
| 116 | + | br := bufio.NewReader(r) | |||
| 117 | + | for { | |||
| 118 | + | line, err := br.ReadBytes('\n') | |||
| 119 | + | if len(line) > 0 { | |||
| 120 | + | line = bytes.TrimRight(line, "\r\n") | |||
| 121 | + | if len(line) > 0 { | |||
| 122 | + | if ferr := fn(line); ferr != nil { | |||
| 123 | + | return ferr | |||
| 124 | + | } | |||
| 125 | + | } | |||
| 126 | + | } | |||
| 127 | + | if err == io.EOF { | |||
| 128 | + | return nil | |||
| 129 | + | } | |||
| 130 | + | if err != nil { | |||
| 131 | + | return err | |||
| 132 | + | } | |||
| 133 | + | } | |||
| 134 | + | } | |||
| 135 | + | ||||
| 113 | // SessionUsage returns the most recent cumulative usage row stored in | 136 | // SessionUsage returns the most recent cumulative usage row stored in | ||
| 114 | // a session file. Sessions append one usage row per completed turn; the | 137 | // a session file. Sessions append one usage row per completed turn; the | ||
| 115 | // latest row's cumulative field is the session total. Missing usage rows | 138 | // latest row's cumulative field is the session total. Missing usage rows | ||
| @@ -122,21 +145,19 @@ func SessionUsage(path string) (provider.Usage, error) { | |||||
| 122 | defer f.Close() | 145 | defer f.Close() | ||
| 123 | 146 | ||||
| 124 | var usage provider.Usage | 147 | var usage provider.Usage | ||
| 125 | - | sc := bufio.NewScanner(f) | 148 | + | if err := forEachJSONLLine(f, func(line []byte) error { |
| 126 | - | sc.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | |||
| 127 | - | for sc.Scan() { | |||
| 128 | var head sessionLineHead | 149 | var head sessionLineHead | ||
| 129 | - | if err := json.Unmarshal(sc.Bytes(), &head); err != nil || head.Type != "usage" { | 150 | + | if err := json.Unmarshal(line, &head); err != nil || head.Type != "usage" { |
| 130 | - | continue | 151 | + | return nil |
| 131 | } | 152 | } | ||
| 132 | var row struct { | 153 | var row struct { | ||
| 133 | Cumulative provider.Usage `json:"cumulative"` | 154 | Cumulative provider.Usage `json:"cumulative"` | ||
| 134 | } | 155 | } | ||
| 135 | - | if err := json.Unmarshal(sc.Bytes(), &row); err == nil { | 156 | + | if err := json.Unmarshal(line, &row); err == nil { |
| 136 | usage = row.Cumulative | 157 | usage = row.Cumulative | ||
| 137 | } | 158 | } | ||
| 138 | - | } | 159 | + | return nil |
| 139 | - | if err := sc.Err(); err != nil { | 160 | + | }); err != nil { |
| 140 | return provider.Usage{}, err | 161 | return provider.Usage{}, err | ||
| 141 | } | 162 | } | ||
| 142 | return usage, nil | 163 | return usage, nil | ||
| @@ -152,28 +173,26 @@ func OpenSession(path string) (*Session, []provider.Message, error) { | |||||
| 152 | 173 | ||||
| 153 | var meta SessionMeta | 174 | var meta SessionMeta | ||
| 154 | var messages []provider.Message | 175 | var messages []provider.Message | ||
| 155 | - | sc := bufio.NewScanner(f) | 176 | + | if err := forEachJSONLLine(f, func(line []byte) error { |
| 156 | - | sc.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | |||
| 157 | - | for sc.Scan() { | |||
| 158 | var head sessionLineHead | 177 | var head sessionLineHead | ||
| 159 | - | if err := json.Unmarshal(sc.Bytes(), &head); err != nil { | 178 | + | if err := json.Unmarshal(line, &head); err != nil { |
| 160 | - | continue | 179 | + | return nil |
| 161 | } | 180 | } | ||
| 162 | switch head.Type { | 181 | switch head.Type { | ||
| 163 | case "meta": | 182 | case "meta": | ||
| 164 | var row struct { | 183 | var row struct { | ||
| 165 | Meta SessionMeta `json:"meta"` | 184 | Meta SessionMeta `json:"meta"` | ||
| 166 | } | 185 | } | ||
| 167 | - | if err := json.Unmarshal(sc.Bytes(), &row); err == nil { | 186 | + | if err := json.Unmarshal(line, &row); err == nil { |
| 168 | meta = row.Meta | 187 | meta = row.Meta | ||
| 169 | } | 188 | } | ||
| 170 | case "message": | 189 | case "message": | ||
| 171 | - | if msg, err := hydrateMessage(sc.Bytes()); err == nil && len(msg.Content) > 0 { | 190 | + | if msg, err := hydrateMessage(line); err == nil && len(msg.Content) > 0 { |
| 172 | messages = append(messages, msg) | 191 | messages = append(messages, msg) | ||
| 173 | } | 192 | } | ||
| 174 | } | 193 | } | ||
| 175 | - | } | 194 | + | return nil |
| 176 | - | if err := sc.Err(); err != nil { | 195 | + | }); err != nil { |
| 177 | return nil, nil, err | 196 | return nil, nil, err | ||
| 178 | } | 197 | } | ||
| 179 | messages = repairToolUseResultPairs(messages) | 198 | messages = repairToolUseResultPairs(messages) | ||
| @@ -333,19 +352,17 @@ func describeSession(path string) SessionSummary { | |||||
| 333 | return s | 352 | return s | ||
| 334 | } | 353 | } | ||
| 335 | defer f.Close() | 354 | defer f.Close() | ||
| 336 | - | sc := bufio.NewScanner(f) | 355 | + | _ = forEachJSONLLine(f, func(line []byte) error { |
| 337 | - | sc.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | |||
| 338 | - | for sc.Scan() { | |||
| 339 | var head sessionLineHead | 356 | var head sessionLineHead | ||
| 340 | - | if err := json.Unmarshal(sc.Bytes(), &head); err != nil { | 357 | + | if err := json.Unmarshal(line, &head); err != nil { |
| 341 | - | continue | 358 | + | return nil |
| 342 | } | 359 | } | ||
| 343 | switch head.Type { | 360 | switch head.Type { | ||
| 344 | case "meta": | 361 | case "meta": | ||
| 345 | var row struct { | 362 | var row struct { | ||
| 346 | Meta SessionMeta `json:"meta"` | 363 | Meta SessionMeta `json:"meta"` | ||
| 347 | } | 364 | } | ||
| 348 | - | if err := json.Unmarshal(sc.Bytes(), &row); err == nil { | 365 | + | if err := json.Unmarshal(line, &row); err == nil { |
| 349 | s.Started = row.Meta.Started | 366 | s.Started = row.Meta.Started | ||
| 350 | s.Model = row.Meta.Model | 367 | s.Model = row.Meta.Model | ||
| 351 | s.Provider = row.Meta.Provider | 368 | s.Provider = row.Meta.Provider | ||
| @@ -354,24 +371,25 @@ func describeSession(path string) SessionSummary { | |||||
| 354 | case "message": | 371 | case "message": | ||
| 355 | s.MessageCount++ | 372 | s.MessageCount++ | ||
| 356 | if s.FirstUserText == "" { | 373 | if s.FirstUserText == "" { | ||
| 357 | - | s.FirstUserText = firstUserText(sc.Bytes()) | 374 | + | s.FirstUserText = firstUserText(line) |
| 358 | } | 375 | } | ||
| 359 | case "rename": | 376 | case "rename": | ||
| 360 | var row struct { | 377 | var row struct { | ||
| 361 | Title string `json:"title"` | 378 | Title string `json:"title"` | ||
| 362 | } | 379 | } | ||
| 363 | - | if err := json.Unmarshal(sc.Bytes(), &row); err == nil && row.Title != "" { | 380 | + | if err := json.Unmarshal(line, &row); err == nil && row.Title != "" { |
| 364 | s.Title = row.Title | 381 | s.Title = row.Title | ||
| 365 | } | 382 | } | ||
| 366 | case "usage": | 383 | case "usage": | ||
| 367 | var row struct { | 384 | var row struct { | ||
| 368 | Cumulative provider.Usage `json:"cumulative"` | 385 | Cumulative provider.Usage `json:"cumulative"` | ||
| 369 | } | 386 | } | ||
| 370 | - | if err := json.Unmarshal(sc.Bytes(), &row); err == nil { | 387 | + | if err := json.Unmarshal(line, &row); err == nil { |
| 371 | s.TotalCost = row.Cumulative.CostUSD | 388 | s.TotalCost = row.Cumulative.CostUSD | ||
| 372 | } | 389 | } | ||
| 373 | } | 390 | } | ||
| 374 | - | } | 391 | + | return nil |
| 392 | + | }) | |||
| 375 | return s | 393 | return s | ||
| 376 | } | 394 | } | ||
| 377 | 395 | ||||
| @@ -429,18 +447,19 @@ func sessionHasNoMessages(path string) bool { | |||||
| 429 | return false | 447 | return false | ||
| 430 | } | 448 | } | ||
| 431 | defer f.Close() | 449 | defer f.Close() | ||
| 432 | - | sc := bufio.NewScanner(f) | 450 | + | hasMessage := false |
| 433 | - | sc.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | 451 | + | _ = forEachJSONLLine(f, func(line []byte) error { |
| 434 | - | for sc.Scan() { | |||
| 435 | var head sessionLineHead | 452 | var head sessionLineHead | ||
| 436 | - | if err := json.Unmarshal(sc.Bytes(), &head); err != nil { | 453 | + | if err := json.Unmarshal(line, &head); err != nil { |
| 437 | - | continue | 454 | + | return nil |
| 438 | } | 455 | } | ||
| 439 | if head.Type == "message" { | 456 | if head.Type == "message" { | ||
| 440 | - | return false | 457 | + | hasMessage = true |
| 458 | + | return io.EOF | |||
| 441 | } | 459 | } | ||
| 442 | - | } | 460 | + | return nil |
| 443 | - | return true | 461 | + | }) |
| 462 | + | return !hasMessage | |||
| 444 | } | 463 | } | ||
| 445 | 464 | ||||
| 446 | // ListSessions returns session file paths for cwd, most-recently- | 465 | // ListSessions returns session file paths for cwd, most-recently- | ||
M internal/core/session_portable.go +84 -78
| @@ -2,6 +2,7 @@ package core | |||
| 2 | 2 | ||
| 3 | 3 | import ( | |
| 4 | 4 | "bufio" | |
| 5 | + | "bytes" | |
| 5 | 6 | "encoding/json" | |
| 6 | 7 | "errors" | |
| 7 | 8 | "fmt" | |
| @@ -66,7 +67,11 @@ func ExportSession(srcPath, dstPath string) (string, error) { | |||
| 66 | 67 | firstPrompt := "" | |
| 67 | 68 | if !strings.HasSuffix(strings.ToLower(dstPath), PortableExt) { | |
| 68 | 69 | if fi, _ := os.Stat(dstPath); fi == nil || fi.IsDir() { | |
| 69 | - | firstPrompt = firstUserPrompt(sc) | |
| 70 | + | p, err := firstUserPrompt(src) | |
| 71 | + | if err != nil { | |
| 72 | + | return "", fmt.Errorf("export: read first prompt: %w", err) | |
| 73 | + | } | |
| 74 | + | firstPrompt = p | |
| 70 | 75 | } | |
| 71 | 76 | } | |
| 72 | 77 | ||
| @@ -110,28 +115,32 @@ func ExportSession(srcPath, dstPath string) (string, error) { | |||
| 110 | 115 | return "", err | |
| 111 | 116 | } | |
| 112 | 117 | ||
| 113 | - | // Stream every non-meta row verbatim. | |
| 114 | - | sc2 := bufio.NewScanner(src) | |
| 115 | - | sc2.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | |
| 116 | - | for sc2.Scan() { | |
| 117 | - | line := sc2.Bytes() | |
| 118 | - | var h sessionLineHead | |
| 119 | - | if err := json.Unmarshal(line, &h); err != nil { | |
| 120 | - | continue | |
| 121 | - | } | |
| 122 | - | if h.Type == "meta" { | |
| 123 | - | continue // already wrote a rewritten copy above | |
| 118 | + | // Stream every non-meta row verbatim. Use ReadBytes instead of | |
| 119 | + | // bufio.Scanner: large sessions can contain very long JSONL rows | |
| 120 | + | // (image blocks, big tool outputs, compacted history) that exceed | |
| 121 | + | // Scanner's token limit and fail with "token too long". | |
| 122 | + | r := bufio.NewReader(src) | |
| 123 | + | for { | |
| 124 | + | line, err := r.ReadBytes('\n') | |
| 125 | + | if len(line) > 0 { | |
| 126 | + | line = bytes.TrimRight(line, "\r\n") | |
| 127 | + | var h sessionLineHead | |
| 128 | + | if err := json.Unmarshal(line, &h); err == nil && h.Type != "meta" { | |
| 129 | + | if _, werr := bw.Write(line); werr != nil { | |
| 130 | + | return "", werr | |
| 131 | + | } | |
| 132 | + | if werr := bw.WriteByte('\n'); werr != nil { | |
| 133 | + | return "", werr | |
| 134 | + | } | |
| 135 | + | } | |
| 124 | 136 | } | |
| 125 | - | if _, err := bw.Write(line); err != nil { | |
| 126 | - | return "", err | |
| 137 | + | if err == io.EOF { | |
| 138 | + | break | |
| 127 | 139 | } | |
| 128 | - | if err := bw.WriteByte('\n'); err != nil { | |
| 129 | - | return "", err | |
| 140 | + | if err != nil { | |
| 141 | + | return "", fmt.Errorf("export: read source: %w", err) | |
| 130 | 142 | } | |
| 131 | 143 | } | |
| 132 | - | if err := sc2.Err(); err != nil { | |
| 133 | - | return "", fmt.Errorf("export: read source: %w", err) | |
| 134 | - | } | |
| 135 | 144 | if err := bw.Flush(); err != nil { | |
| 136 | 145 | return "", err | |
| 137 | 146 | } | |
| @@ -207,29 +216,22 @@ func ImportSession(srcPath, root, cwd, version string) (string, error) { | |||
| 207 | 216 | return "", err | |
| 208 | 217 | } | |
| 209 | 218 | ||
| 210 | - | // Rewind the source and stream every non-meta row. | |
| 219 | + | // Rewind the source and stream every non-meta row. Avoid | |
| 220 | + | // bufio.Scanner so exported sessions with huge JSONL rows import | |
| 221 | + | // cleanly. | |
| 211 | 222 | if _, err := src.Seek(0, io.SeekStart); err != nil { | |
| 212 | 223 | return "", fmt.Errorf("import: rewind: %w", err) | |
| 213 | 224 | } | |
| 214 | - | sc2 := bufio.NewScanner(src) | |
| 215 | - | sc2.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | |
| 216 | - | for sc2.Scan() { | |
| 217 | - | line := sc2.Bytes() | |
| 225 | + | if err := forEachJSONLLine(src, func(line []byte) error { | |
| 218 | 226 | var h sessionLineHead | |
| 219 | - | if err := json.Unmarshal(line, &h); err != nil { | |
| 220 | - | continue | |
| 221 | - | } | |
| 222 | - | if h.Type == "meta" { | |
| 223 | - | continue | |
| 227 | + | if err := json.Unmarshal(line, &h); err != nil || h.Type == "meta" { | |
| 228 | + | return nil | |
| 224 | 229 | } | |
| 225 | 230 | if _, err := bw.Write(line); err != nil { | |
| 226 | - | return "", err | |
| 231 | + | return err | |
| 227 | 232 | } | |
| 228 | - | if err := bw.WriteByte('\n'); err != nil { | |
| 229 | - | return "", err | |
| 230 | - | } | |
| 231 | - | } | |
| 232 | - | if err := sc2.Err(); err != nil { | |
| 233 | + | return bw.WriteByte('\n') | |
| 234 | + | }); err != nil { | |
| 233 | 235 | return "", fmt.Errorf("import: read source: %w", err) | |
| 234 | 236 | } | |
| 235 | 237 | if err := bw.Flush(); err != nil { | |
| @@ -321,34 +323,39 @@ func BranchSession(parentPath, root, cwd, version string, upToMessageIdx int) (s | |||
| 321 | 323 | ||
| 322 | 324 | // Copy message rows up to the cut point, plus all usage rows | |
| 323 | 325 | // that land before the cut (they describe the cost of those | |
| 324 | - | // messages). | |
| 326 | + | // messages). Rewind and use the large-row-safe JSONL reader. | |
| 327 | + | if _, err := src.Seek(0, io.SeekStart); err != nil { | |
| 328 | + | return "", fmt.Errorf("branch: rewind parent: %w", err) | |
| 329 | + | } | |
| 325 | 330 | msgCount := 0 | |
| 326 | - | for sc.Scan() && msgCount < upToMessageIdx { | |
| 327 | - | line := sc.Bytes() | |
| 331 | + | if err := forEachJSONLLine(src, func(line []byte) error { | |
| 332 | + | if msgCount >= upToMessageIdx { | |
| 333 | + | return io.EOF | |
| 334 | + | } | |
| 328 | 335 | var h sessionLineHead | |
| 329 | 336 | if err := json.Unmarshal(line, &h); err != nil { | |
| 330 | - | continue | |
| 337 | + | return nil | |
| 331 | 338 | } | |
| 332 | 339 | switch h.Type { | |
| 333 | 340 | case "message": | |
| 334 | 341 | if _, err := bw.Write(line); err != nil { | |
| 335 | - | return "", err | |
| 342 | + | return err | |
| 336 | 343 | } | |
| 337 | 344 | if err := bw.WriteByte('\n'); err != nil { | |
| 338 | - | return "", err | |
| 345 | + | return err | |
| 339 | 346 | } | |
| 340 | 347 | msgCount++ | |
| 341 | 348 | case "usage": | |
| 342 | 349 | if _, err := bw.Write(line); err != nil { | |
| 343 | - | return "", err | |
| 350 | + | return err | |
| 344 | 351 | } | |
| 345 | 352 | if err := bw.WriteByte('\n'); err != nil { | |
| 346 | - | return "", err | |
| 353 | + | return err | |
| 347 | 354 | } | |
| 348 | 355 | // don't increment msgCount for usage rows | |
| 349 | 356 | } | |
| 350 | - | } | |
| 351 | - | if err := sc.Err(); err != nil { | |
| 357 | + | return nil | |
| 358 | + | }); err != nil && err != io.EOF { | |
| 352 | 359 | return "", fmt.Errorf("branch: read parent: %w", err) | |
| 353 | 360 | } | |
| 354 | 361 | if err := bw.Flush(); err != nil { | |
| @@ -456,43 +463,42 @@ func FindSessionByID(root, cwd, id string) string { | |||
| 456 | 463 | return "" | |
| 457 | 464 | } | |
| 458 | 465 | ||
| 459 | - | // firstUserPrompt scans forward from the current scanner position | |
| 460 | - | // looking for the first user-role message and returns its text | |
| 461 | - | // (trimmed, short). Used to build a humane export filename. | |
| 462 | - | func firstUserPrompt(sc *bufio.Scanner) string { | |
| 463 | - | for sc.Scan() { | |
| 464 | - | var line sessionLine | |
| 465 | - | if err := json.Unmarshal(sc.Bytes(), &line); err != nil { | |
| 466 | - | continue | |
| 467 | - | } | |
| 468 | - | if line.Type != "message" || line.Message == nil || line.Message.Role != "user" { | |
| 469 | - | continue | |
| 470 | - | } | |
| 471 | - | for _, c := range line.Message.Content { | |
| 472 | - | // Use type name to avoid an import of provider here beyond | |
| 473 | - | // the already-imported alias; TextBlock is the only content | |
| 474 | - | // shape that yields a useful preview, so we just look for | |
| 475 | - | // something with a reasonable string form. | |
| 476 | - | s := fmt.Sprintf("%v", c) | |
| 477 | - | _ = s // formatted value is too noisy; go straight to typed path | |
| 466 | + | // firstUserPrompt scans forward from the current source position | |
| 467 | + | // looking for the first user-role message and returns its text. | |
| 468 | + | // Used to build a humane export filename. Uses Reader instead of | |
| 469 | + | // Scanner so a very large JSONL row before the first user prompt | |
| 470 | + | // cannot trip Scanner's token limit. | |
| 471 | + | func firstUserPrompt(src io.Reader) (string, error) { | |
| 472 | + | r := bufio.NewReader(src) | |
| 473 | + | for { | |
| 474 | + | lineBytes, err := r.ReadBytes('\n') | |
| 475 | + | if len(lineBytes) > 0 { | |
| 476 | + | lineBytes = bytes.TrimRight(lineBytes, "\r\n") | |
| 477 | + | var line sessionLine | |
| 478 | + | if err := json.Unmarshal(lineBytes, &line); err == nil { | |
| 479 | + | if line.Type == "message" && line.Message != nil && line.Message.Role == "user" { | |
| 480 | + | b, _ := json.Marshal(line.Message) | |
| 481 | + | var m struct { | |
| 482 | + | Content []struct { | |
| 483 | + | Text string `json:"text"` | |
| 484 | + | } `json:"content"` | |
| 485 | + | } | |
| 486 | + | _ = json.Unmarshal(b, &m) | |
| 487 | + | for _, c := range m.Content { | |
| 488 | + | if c.Text != "" { | |
| 489 | + | return c.Text, nil | |
| 490 | + | } | |
| 491 | + | } | |
| 492 | + | } | |
| 493 | + | } | |
| 478 | 494 | } | |
| 479 | - | // Simpler: marshal the message and fish out the first "text" | |
| 480 | - | // we can find. Avoids reaching into the provider package just | |
| 481 | - | // for an interface type check. | |
| 482 | - | b, _ := json.Marshal(line.Message) | |
| 483 | - | var m struct { | |
| 484 | - | Content []struct { | |
| 485 | - | Text string `json:"text"` | |
| 486 | - | } `json:"content"` | |
| 495 | + | if err == io.EOF { | |
| 496 | + | return "", nil | |
| 487 | 497 | } | |
| 488 | - | _ = json.Unmarshal(b, &m) | |
| 489 | - | for _, c := range m.Content { | |
| 490 | - | if c.Text != "" { | |
| 491 | - | return c.Text | |
| 492 | - | } | |
| 498 | + | if err != nil { | |
| 499 | + | return "", err | |
| 493 | 500 | } | |
| 494 | 501 | } | |
| 495 | - | return "" | |
| 496 | 502 | } | |
| 497 | 503 | ||
| 498 | 504 | // filenameFor builds a descriptive .zotsession filename from the | |
| @@ -2,6 +2,7 @@ package core | |||||
| 2 | 2 | ||||
| 3 | import ( | 3 | import ( | ||
| 4 | "bufio" | 4 | "bufio" | ||
| 5 | + | "bytes" | |||
| 5 | "encoding/json" | 6 | "encoding/json" | ||
| 6 | "errors" | 7 | "errors" | ||
| 7 | "fmt" | 8 | "fmt" | ||
| @@ -66,7 +67,11 @@ func ExportSession(srcPath, dstPath string) (string, error) { | |||||
| 66 | firstPrompt := "" | 67 | firstPrompt := "" | ||
| 67 | if !strings.HasSuffix(strings.ToLower(dstPath), PortableExt) { | 68 | if !strings.HasSuffix(strings.ToLower(dstPath), PortableExt) { | ||
| 68 | if fi, _ := os.Stat(dstPath); fi == nil || fi.IsDir() { | 69 | if fi, _ := os.Stat(dstPath); fi == nil || fi.IsDir() { | ||
| 69 | - | firstPrompt = firstUserPrompt(sc) | 70 | + | p, err := firstUserPrompt(src) |
| 71 | + | if err != nil { | |||
| 72 | + | return "", fmt.Errorf("export: read first prompt: %w", err) | |||
| 73 | + | } | |||
| 74 | + | firstPrompt = p | |||
| 70 | } | 75 | } | ||
| 71 | } | 76 | } | ||
| 72 | 77 | ||||
| @@ -110,28 +115,32 @@ func ExportSession(srcPath, dstPath string) (string, error) { | |||||
| 110 | return "", err | 115 | return "", err | ||
| 111 | } | 116 | } | ||
| 112 | 117 | ||||
| 113 | - | // Stream every non-meta row verbatim. | 118 | + | // Stream every non-meta row verbatim. Use ReadBytes instead of |
| 114 | - | sc2 := bufio.NewScanner(src) | 119 | + | // bufio.Scanner: large sessions can contain very long JSONL rows |
| 115 | - | sc2.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | 120 | + | // (image blocks, big tool outputs, compacted history) that exceed |
| 116 | - | for sc2.Scan() { | 121 | + | // Scanner's token limit and fail with "token too long". |
| 117 | - | line := sc2.Bytes() | 122 | + | r := bufio.NewReader(src) |
| 118 | - | var h sessionLineHead | 123 | + | for { |
| 119 | - | if err := json.Unmarshal(line, &h); err != nil { | 124 | + | line, err := r.ReadBytes('\n') |
| 120 | - | continue | 125 | + | if len(line) > 0 { |
| 121 | - | } | 126 | + | line = bytes.TrimRight(line, "\r\n") |
| 122 | - | if h.Type == "meta" { | 127 | + | var h sessionLineHead |
| 123 | - | continue // already wrote a rewritten copy above | 128 | + | if err := json.Unmarshal(line, &h); err == nil && h.Type != "meta" { |
| 129 | + | if _, werr := bw.Write(line); werr != nil { | |||
| 130 | + | return "", werr | |||
| 131 | + | } | |||
| 132 | + | if werr := bw.WriteByte('\n'); werr != nil { | |||
| 133 | + | return "", werr | |||
| 134 | + | } | |||
| 135 | + | } | |||
| 124 | } | 136 | } | ||
| 125 | - | if _, err := bw.Write(line); err != nil { | 137 | + | if err == io.EOF { |
| 126 | - | return "", err | 138 | + | break |
| 127 | } | 139 | } | ||
| 128 | - | if err := bw.WriteByte('\n'); err != nil { | 140 | + | if err != nil { |
| 129 | - | return "", err | 141 | + | return "", fmt.Errorf("export: read source: %w", err) |
| 130 | } | 142 | } | ||
| 131 | } | 143 | } | ||
| 132 | - | if err := sc2.Err(); err != nil { | |||
| 133 | - | return "", fmt.Errorf("export: read source: %w", err) | |||
| 134 | - | } | |||
| 135 | if err := bw.Flush(); err != nil { | 144 | if err := bw.Flush(); err != nil { | ||
| 136 | return "", err | 145 | return "", err | ||
| 137 | } | 146 | } | ||
| @@ -207,29 +216,22 @@ func ImportSession(srcPath, root, cwd, version string) (string, error) { | |||||
| 207 | return "", err | 216 | return "", err | ||
| 208 | } | 217 | } | ||
| 209 | 218 | ||||
| 210 | - | // Rewind the source and stream every non-meta row. | 219 | + | // Rewind the source and stream every non-meta row. Avoid |
| 220 | + | // bufio.Scanner so exported sessions with huge JSONL rows import | |||
| 221 | + | // cleanly. | |||
| 211 | if _, err := src.Seek(0, io.SeekStart); err != nil { | 222 | if _, err := src.Seek(0, io.SeekStart); err != nil { | ||
| 212 | return "", fmt.Errorf("import: rewind: %w", err) | 223 | return "", fmt.Errorf("import: rewind: %w", err) | ||
| 213 | } | 224 | } | ||
| 214 | - | sc2 := bufio.NewScanner(src) | 225 | + | if err := forEachJSONLLine(src, func(line []byte) error { |
| 215 | - | sc2.Buffer(make([]byte, 0, 64*1024), 20*1024*1024) | |||
| 216 | - | for sc2.Scan() { | |||
| 217 | - | line := sc2.Bytes() | |||
| 218 | var h sessionLineHead | 226 | var h sessionLineHead | ||
| 219 | - | if err := json.Unmarshal(line, &h); err != nil { | 227 | + | if err := json.Unmarshal(line, &h); err != nil || h.Type == "meta" { |
| 220 | - | continue | 228 | + | return nil |
| 221 | - | } | |||
| 222 | - | if h.Type == "meta" { | |||
| 223 | - | continue | |||
| 224 | } | 229 | } | ||
| 225 | if _, err := bw.Write(line); err != nil { | 230 | if _, err := bw.Write(line); err != nil { | ||
| 226 | - | return "", err | 231 | + | return err |
| 227 | } | 232 | } | ||
| 228 | - | if err := bw.WriteByte('\n'); err != nil { | 233 | + | return bw.WriteByte('\n') |
| 229 | - | return "", err | 234 | + | }); err != nil { |
| 230 | - | } | |||
| 231 | - | } | |||
| 232 | - | if err := sc2.Err(); err != nil { | |||
| 233 | return "", fmt.Errorf("import: read source: %w", err) | 235 | return "", fmt.Errorf("import: read source: %w", err) | ||
| 234 | } | 236 | } | ||
| 235 | if err := bw.Flush(); err != nil { | 237 | if err := bw.Flush(); err != nil { | ||
| @@ -321,34 +323,39 @@ func BranchSession(parentPath, root, cwd, version string, upToMessageIdx int) (s | |||||
| 321 | 323 | ||||
| 322 | // Copy message rows up to the cut point, plus all usage rows | 324 | // Copy message rows up to the cut point, plus all usage rows | ||
| 323 | // that land before the cut (they describe the cost of those | 325 | // that land before the cut (they describe the cost of those | ||
| 324 | - | // messages). | 326 | + | // messages). Rewind and use the large-row-safe JSONL reader. |
| 327 | + | if _, err := src.Seek(0, io.SeekStart); err != nil { | |||
| 328 | + | return "", fmt.Errorf("branch: rewind parent: %w", err) | |||
| 329 | + | } | |||
| 325 | msgCount := 0 | 330 | msgCount := 0 | ||
| 326 | - | for sc.Scan() && msgCount < upToMessageIdx { | 331 | + | if err := forEachJSONLLine(src, func(line []byte) error { |
| 327 | - | line := sc.Bytes() | 332 | + | if msgCount >= upToMessageIdx { |
| 333 | + | return io.EOF | |||
| 334 | + | } | |||
| 328 | var h sessionLineHead | 335 | var h sessionLineHead | ||
| 329 | if err := json.Unmarshal(line, &h); err != nil { | 336 | if err := json.Unmarshal(line, &h); err != nil { | ||
| 330 | - | continue | 337 | + | return nil |
| 331 | } | 338 | } | ||
| 332 | switch h.Type { | 339 | switch h.Type { | ||
| 333 | case "message": | 340 | case "message": | ||
| 334 | if _, err := bw.Write(line); err != nil { | 341 | if _, err := bw.Write(line); err != nil { | ||
| 335 | - | return "", err | 342 | + | return err |
| 336 | } | 343 | } | ||
| 337 | if err := bw.WriteByte('\n'); err != nil { | 344 | if err := bw.WriteByte('\n'); err != nil { | ||
| 338 | - | return "", err | 345 | + | return err |
| 339 | } | 346 | } | ||
| 340 | msgCount++ | 347 | msgCount++ | ||
| 341 | case "usage": | 348 | case "usage": | ||
| 342 | if _, err := bw.Write(line); err != nil { | 349 | if _, err := bw.Write(line); err != nil { | ||
| 343 | - | return "", err | 350 | + | return err |
| 344 | } | 351 | } | ||
| 345 | if err := bw.WriteByte('\n'); err != nil { | 352 | if err := bw.WriteByte('\n'); err != nil { | ||
| 346 | - | return "", err | 353 | + | return err |
| 347 | } | 354 | } | ||
| 348 | // don't increment msgCount for usage rows | 355 | // don't increment msgCount for usage rows | ||
| 349 | } | 356 | } | ||
| 350 | - | } | 357 | + | return nil |
| 351 | - | if err := sc.Err(); err != nil { | 358 | + | }); err != nil && err != io.EOF { |
| 352 | return "", fmt.Errorf("branch: read parent: %w", err) | 359 | return "", fmt.Errorf("branch: read parent: %w", err) | ||
| 353 | } | 360 | } | ||
| 354 | if err := bw.Flush(); err != nil { | 361 | if err := bw.Flush(); err != nil { | ||
| @@ -456,43 +463,42 @@ func FindSessionByID(root, cwd, id string) string { | |||||
| 456 | return "" | 463 | return "" | ||
| 457 | } | 464 | } | ||
| 458 | 465 | ||||
| 459 | - | // firstUserPrompt scans forward from the current scanner position | 466 | + | // firstUserPrompt scans forward from the current source position |
| 460 | - | // looking for the first user-role message and returns its text | 467 | + | // looking for the first user-role message and returns its text. |
| 461 | - | // (trimmed, short). Used to build a humane export filename. | 468 | + | // Used to build a humane export filename. Uses Reader instead of |
| 462 | - | func firstUserPrompt(sc *bufio.Scanner) string { | 469 | + | // Scanner so a very large JSONL row before the first user prompt |
| 463 | - | for sc.Scan() { | 470 | + | // cannot trip Scanner's token limit. |
| 464 | - | var line sessionLine | 471 | + | func firstUserPrompt(src io.Reader) (string, error) { |
| 465 | - | if err := json.Unmarshal(sc.Bytes(), &line); err != nil { | 472 | + | r := bufio.NewReader(src) |
| 466 | - | continue | 473 | + | for { |
| 467 | - | } | 474 | + | lineBytes, err := r.ReadBytes('\n') |
| 468 | - | if line.Type != "message" || line.Message == nil || line.Message.Role != "user" { | 475 | + | if len(lineBytes) > 0 { |
| 469 | - | continue | 476 | + | lineBytes = bytes.TrimRight(lineBytes, "\r\n") |
| 470 | - | } | 477 | + | var line sessionLine |
| 471 | - | for _, c := range line.Message.Content { | 478 | + | if err := json.Unmarshal(lineBytes, &line); err == nil { |
| 472 | - | // Use type name to avoid an import of provider here beyond | 479 | + | if line.Type == "message" && line.Message != nil && line.Message.Role == "user" { |
| 473 | - | // the already-imported alias; TextBlock is the only content | 480 | + | b, _ := json.Marshal(line.Message) |
| 474 | - | // shape that yields a useful preview, so we just look for | 481 | + | var m struct { |
| 475 | - | // something with a reasonable string form. | 482 | + | Content []struct { |
| 476 | - | s := fmt.Sprintf("%v", c) | 483 | + | Text string `json:"text"` |
| 477 | - | _ = s // formatted value is too noisy; go straight to typed path | 484 | + | } `json:"content"` |
| 485 | + | } | |||
| 486 | + | _ = json.Unmarshal(b, &m) | |||
| 487 | + | for _, c := range m.Content { | |||
| 488 | + | if c.Text != "" { | |||
| 489 | + | return c.Text, nil | |||
| 490 | + | } | |||
| 491 | + | } | |||
| 492 | + | } | |||
| 493 | + | } | |||
| 478 | } | 494 | } | ||
| 479 | - | // Simpler: marshal the message and fish out the first "text" | 495 | + | if err == io.EOF { |
| 480 | - | // we can find. Avoids reaching into the provider package just | 496 | + | return "", nil |
| 481 | - | // for an interface type check. | |||
| 482 | - | b, _ := json.Marshal(line.Message) | |||
| 483 | - | var m struct { | |||
| 484 | - | Content []struct { | |||
| 485 | - | Text string `json:"text"` | |||
| 486 | - | } `json:"content"` | |||
| 487 | } | 497 | } | ||
| 488 | - | _ = json.Unmarshal(b, &m) | 498 | + | if err != nil { |
| 489 | - | for _, c := range m.Content { | 499 | + | return "", err |
| 490 | - | if c.Text != "" { | |||
| 491 | - | return c.Text | |||
| 492 | - | } | |||
| 493 | } | 500 | } | ||
| 494 | } | 501 | } | ||
| 495 | - | return "" | |||
| 496 | } | 502 | } | ||
| 497 | 503 | ||||
| 498 | // filenameFor builds a descriptive .zotsession filename from the | 504 | // filenameFor builds a descriptive .zotsession filename from the | ||
M internal/core/session_portable_test.go +50 -0
| @@ -1,6 +1,7 @@ | |||
| 1 | 1 | package core | |
| 2 | 2 | ||
| 3 | 3 | import ( | |
| 4 | + | "encoding/json" | |
| 4 | 5 | "os" | |
| 5 | 6 | "path/filepath" | |
| 6 | 7 | "strings" | |
| @@ -109,6 +110,55 @@ func TestExportToFilePath(t *testing.T) { | |||
| 109 | 110 | ||
| 110 | 111 | // TestExportStripsCWDFromMeta verifies the exported meta no longer | |
| 111 | 112 | // carries the source user's cwd (not useful to the recipient). | |
| 113 | + | func TestExportSessionHandlesHugeJSONLRows(t *testing.T) { | |
| 114 | + | root := t.TempDir() | |
| 115 | + | sess, err := NewSession(root, "/cwd", "anthropic", "claude-opus-4-7", "0.0.0-test") | |
| 116 | + | if err != nil { | |
| 117 | + | t.Fatal(err) | |
| 118 | + | } | |
| 119 | + | _ = sess.AppendMessage(provider.Message{ | |
| 120 | + | Role: provider.RoleUser, | |
| 121 | + | Content: []provider.Content{provider.TextBlock{Text: "first prompt names export"}}, | |
| 122 | + | }) | |
| 123 | + | _ = sess.AppendMessage(provider.Message{ | |
| 124 | + | Role: provider.RoleAssistant, | |
| 125 | + | Content: []provider.Content{provider.TextBlock{Text: strings.Repeat("x", 22*1024*1024)}}, | |
| 126 | + | }) | |
| 127 | + | _ = sess.Close() | |
| 128 | + | ||
| 129 | + | exportDir := t.TempDir() | |
| 130 | + | out, err := ExportSession(sess.Path, exportDir) | |
| 131 | + | if err != nil { | |
| 132 | + | t.Fatalf("ExportSession with huge row: %v", err) | |
| 133 | + | } | |
| 134 | + | if _, err := os.Stat(out); err != nil { | |
| 135 | + | t.Fatalf("exported file missing: %v", err) | |
| 136 | + | } | |
| 137 | + | opened, msgs, err := OpenSession(sess.Path) | |
| 138 | + | if err != nil { | |
| 139 | + | t.Fatalf("OpenSession with huge row: %v", err) | |
| 140 | + | } | |
| 141 | + | _ = opened.Close() | |
| 142 | + | if len(msgs) != 2 { | |
| 143 | + | t.Fatalf("OpenSession messages=%d, want 2", len(msgs)) | |
| 144 | + | } | |
| 145 | + | // Ensure the exported file is still readable JSONL and contains the | |
| 146 | + | // huge assistant message row, not just the meta/header. | |
| 147 | + | b, err := os.ReadFile(out) | |
| 148 | + | if err != nil { | |
| 149 | + | t.Fatal(err) | |
| 150 | + | } | |
| 151 | + | if !strings.Contains(string(b), strings.Repeat("x", 1024)) { | |
| 152 | + | t.Fatalf("export does not appear to contain huge assistant row") | |
| 153 | + | } | |
| 154 | + | for n, line := range strings.Split(strings.TrimSpace(string(b)), "\n") { | |
| 155 | + | var raw map[string]any | |
| 156 | + | if err := json.Unmarshal([]byte(line), &raw); err != nil { | |
| 157 | + | t.Fatalf("exported line %d is invalid json: %v", n+1, err) | |
| 158 | + | } | |
| 159 | + | } | |
| 160 | + | } | |
| 161 | + | ||
| 112 | 162 | func TestExportStripsCWDFromMeta(t *testing.T) { | |
| 113 | 163 | root := t.TempDir() | |
| 114 | 164 | sess, err := NewSession(root, "/original/cwd", "anthropic", "claude-opus-4-7", "0.0.0-test") | |
| @@ -1,6 +1,7 @@ | |||||
| 1 | package core | 1 | package core | ||
| 2 | 2 | ||||
| 3 | import ( | 3 | import ( | ||
| 4 | + | "encoding/json" | |||
| 4 | "os" | 5 | "os" | ||
| 5 | "path/filepath" | 6 | "path/filepath" | ||
| 6 | "strings" | 7 | "strings" | ||
| @@ -109,6 +110,55 @@ func TestExportToFilePath(t *testing.T) { | |||||
| 109 | 110 | ||||
| 110 | // TestExportStripsCWDFromMeta verifies the exported meta no longer | 111 | // TestExportStripsCWDFromMeta verifies the exported meta no longer | ||
| 111 | // carries the source user's cwd (not useful to the recipient). | 112 | // carries the source user's cwd (not useful to the recipient). | ||
| 113 | + | func TestExportSessionHandlesHugeJSONLRows(t *testing.T) { | |||
| 114 | + | root := t.TempDir() | |||
| 115 | + | sess, err := NewSession(root, "/cwd", "anthropic", "claude-opus-4-7", "0.0.0-test") | |||
| 116 | + | if err != nil { | |||
| 117 | + | t.Fatal(err) | |||
| 118 | + | } | |||
| 119 | + | _ = sess.AppendMessage(provider.Message{ | |||
| 120 | + | Role: provider.RoleUser, | |||
| 121 | + | Content: []provider.Content{provider.TextBlock{Text: "first prompt names export"}}, | |||
| 122 | + | }) | |||
| 123 | + | _ = sess.AppendMessage(provider.Message{ | |||
| 124 | + | Role: provider.RoleAssistant, | |||
| 125 | + | Content: []provider.Content{provider.TextBlock{Text: strings.Repeat("x", 22*1024*1024)}}, | |||
| 126 | + | }) | |||
| 127 | + | _ = sess.Close() | |||
| 128 | + | ||||
| 129 | + | exportDir := t.TempDir() | |||
| 130 | + | out, err := ExportSession(sess.Path, exportDir) | |||
| 131 | + | if err != nil { | |||
| 132 | + | t.Fatalf("ExportSession with huge row: %v", err) | |||
| 133 | + | } | |||
| 134 | + | if _, err := os.Stat(out); err != nil { | |||
| 135 | + | t.Fatalf("exported file missing: %v", err) | |||
| 136 | + | } | |||
| 137 | + | opened, msgs, err := OpenSession(sess.Path) | |||
| 138 | + | if err != nil { | |||
| 139 | + | t.Fatalf("OpenSession with huge row: %v", err) | |||
| 140 | + | } | |||
| 141 | + | _ = opened.Close() | |||
| 142 | + | if len(msgs) != 2 { | |||
| 143 | + | t.Fatalf("OpenSession messages=%d, want 2", len(msgs)) | |||
| 144 | + | } | |||
| 145 | + | // Ensure the exported file is still readable JSONL and contains the | |||
| 146 | + | // huge assistant message row, not just the meta/header. | |||
| 147 | + | b, err := os.ReadFile(out) | |||
| 148 | + | if err != nil { | |||
| 149 | + | t.Fatal(err) | |||
| 150 | + | } | |||
| 151 | + | if !strings.Contains(string(b), strings.Repeat("x", 1024)) { | |||
| 152 | + | t.Fatalf("export does not appear to contain huge assistant row") | |||
| 153 | + | } | |||
| 154 | + | for n, line := range strings.Split(strings.TrimSpace(string(b)), "\n") { | |||
| 155 | + | var raw map[string]any | |||
| 156 | + | if err := json.Unmarshal([]byte(line), &raw); err != nil { | |||
| 157 | + | t.Fatalf("exported line %d is invalid json: %v", n+1, err) | |||
| 158 | + | } | |||
| 159 | + | } | |||
| 160 | + | } | |||
| 161 | + | ||||
| 112 | func TestExportStripsCWDFromMeta(t *testing.T) { | 162 | func TestExportStripsCWDFromMeta(t *testing.T) { | ||
| 113 | root := t.TempDir() | 163 | root := t.TempDir() | ||
| 114 | sess, err := NewSession(root, "/original/cwd", "anthropic", "claude-opus-4-7", "0.0.0-test") | 164 | sess, err := NewSession(root, "/original/cwd", "anthropic", "claude-opus-4-7", "0.0.0-test") | ||