Skip to content

Commit 1b78a5c

Browse files
sfc-gh-pgaursfc-gh-pgaur
committed
Update flow digram and layout digram to use ASCII and not unicode characters
1 parent 48fd8fc commit 1b78a5c

File tree

1 file changed

+103
-103
lines changed
  • cpp/src/arrow/util/alp

1 file changed

+103
-103
lines changed

‎cpp/src/arrow/util/alp/Alp.h‎

Lines changed: 103 additions & 103 deletions
Original file line numberDiff line numberDiff line change
@@ -19,88 +19,88 @@ namespace alp {
1919
* reference and bitpacked. Every exception, where the conversion/reconversion changes the value of
2020
* the float, is stored separately and has to be patched into the decompressed vector afterwards.
2121
*
22-
* ═══════════════════════════════════════════════════════════════════════════════════════
22+
* ==========================================================================================
2323
* ALP COMPRESSION/DECOMPRESSION PIPELINE
24-
* ═══════════════════════════════════════════════════════════════════════════════════════
24+
* ==========================================================================================
2525
*
2626
* COMPRESSION FLOW:
27-
* ─────────────────
27+
* -----------------
2828
*
2929
* Input: float/double array
30-
* │
31-
* ▼
32-
* ┌─────────────────────────────────────────────────────────────────┐
33-
* │ 1. SAMPLING & PRESET GENERATION │
34-
* │ • Sample vectors from dataset │
35-
* │ • Try all exponent/factor combinations (e, f) │
36-
* │ • Select best k combinations for preset │
37-
* └──────────────────────────────────────┬──────────────────────────┘
38-
* │ preset.combinations
39-
* ▼
40-
* ┌─────────────────────────────────────────────────────────────────┐
41-
* │ 2. PER-VECTOR COMPRESSION │
42-
* │ a) Find best (e,f) from preset for this vector │
43-
* │ b) Encode: encoded[i] = int64(value[i] * 10^e * 10^-f) │
44-
* │ c) Verify: if decode(encoded[i]) ≠ value[i] → exception │
45-
* │ d) Replace exceptions with placeholder value │
46-
* └──────────────────────────────────────┬──────────────────────────┘
47-
* │ encoded integers + exceptions
48-
* ▼
49-
* ┌─────────────────────────────────────────────────────────────────┐
50-
* │ 3. FRAME OF REFERENCE (FOR) │
51-
* │ • Find min value in encoded integers │
52-
* │ • Subtract min from all values: delta[i] = encoded[i] - min │
53-
* └──────────────────────────────────────┬──────────────────────────┘
54-
* │ delta values (smaller range)
55-
* ▼
56-
* ┌─────────────────────────────────────────────────────────────────┐
57-
* │ 4. BIT PACKING │
58-
* │ • Calculate bitWidth = log2(max_delta) │
59-
* │ • Pack each value into bitWidth bits │
60-
* │ • Result: tightly packed binary data │
61-
* └──────────────────────────────────────┬──────────────────────────┘
62-
* │ packed bytes
63-
* ▼
64-
* ┌─────────────────────────────────────────────────────────────────┐
65-
* │ 5. SERIALIZATION (see AlpEncodedVector diagram below) │
66-
* │ [VectorInfo][PackedData][ExceptionPos][ExceptionValues] │
67-
* └─────────────────────────────────────────────────────────────────┘
30+
* |
31+
* v
32+
* +------------------------------------------------------------------+
33+
* | 1. SAMPLING & PRESET GENERATION |
34+
* | * Sample vectors from dataset |
35+
* | * Try all exponent/factor combinations (e, f) |
36+
* | * Select best k combinations for preset |
37+
* +------------------------------------+-----------------------------+
38+
* | preset.combinations
39+
* v
40+
* +------------------------------------------------------------------+
41+
* | 2. PER-VECTOR COMPRESSION |
42+
* | a) Find best (e,f) from preset for this vector |
43+
* | b) Encode: encoded[i] = int64(value[i] * 10^e * 10^-f) |
44+
* | c) Verify: if decode(encoded[i]) != value[i] -> exception |
45+
* | d) Replace exceptions with placeholder value |
46+
* +------------------------------------+-----------------------------+
47+
* | encoded integers + exceptions
48+
* v
49+
* +------------------------------------------------------------------+
50+
* | 3. FRAME OF REFERENCE (FOR) |
51+
* | * Find min value in encoded integers |
52+
* | * Subtract min from all values: delta[i] = encoded[i] - min |
53+
* +------------------------------------+-----------------------------+
54+
* | delta values (smaller range)
55+
* v
56+
* +------------------------------------------------------------------+
57+
* | 4. BIT PACKING |
58+
* | * Calculate bitWidth = log2(max_delta) |
59+
* | * Pack each value into bitWidth bits |
60+
* | * Result: tightly packed binary data |
61+
* +------------------------------------+-----------------------------+
62+
* | packed bytes
63+
* v
64+
* +------------------------------------------------------------------+
65+
* | 5. SERIALIZATION (see AlpEncodedVector diagram below) |
66+
* | [VectorInfo][PackedData][ExceptionPos][ExceptionValues] |
67+
* +------------------------------------------------------------------+
6868
*
6969
*
7070
* DECOMPRESSION FLOW:
71-
* ───────────────────
71+
* -------------------
7272
*
73-
* Serialized bytes → AlpEncodedVector::load()
74-
* │
75-
* ▼
76-
* ┌─────────────────────────────────────────────────────────────────┐
77-
* │ 1. BIT UNPACKING │
78-
* │ • Extract bitWidth from metadata │
79-
* │ • Unpack each value from bitWidth bits → delta values │
80-
* └──────────────────────────────────────┬──────────────────────────┘
81-
* │ delta values
82-
* ▼
83-
* ┌─────────────────────────────────────────────────────────────────┐
84-
* │ 2. REVERSE FRAME OF REFERENCE (unFOR) │
85-
* │ • Add back min: encoded[i] = delta[i] + frameOfReference │
86-
* └──────────────────────────────────────┬──────────────────────────┘
87-
* │ encoded integers
88-
* ▼
89-
* ┌─────────────────────────────────────────────────────────────────┐
90-
* │ 3. DECODE │
91-
* │ • Apply inverse formula: value[i] = encoded[i] * 10^-e * 10^f│
92-
* └──────────────────────────────────────┬──────────────────────────┘
93-
* │ decoded floats (with placeholders)
94-
* ▼
95-
* ┌─────────────────────────────────────────────────────────────────┐
96-
* │ 4. PATCH EXCEPTIONS │
97-
* │ • Replace values at exceptionPositions[] with exceptions[] │
98-
* └──────────────────────────────────────┬──────────────────────────┘
99-
* │
100-
* ▼
73+
* Serialized bytes -> AlpEncodedVector::load()
74+
* |
75+
* v
76+
* +------------------------------------------------------------------+
77+
* | 1. BIT UNPACKING |
78+
* | * Extract bitWidth from metadata |
79+
* | * Unpack each value from bitWidth bits -> delta values |
80+
* +------------------------------------+-----------------------------+
81+
* | delta values
82+
* v
83+
* +------------------------------------------------------------------+
84+
* | 2. REVERSE FRAME OF REFERENCE (unFOR) |
85+
* | * Add back min: encoded[i] = delta[i] + frameOfReference |
86+
* +------------------------------------+-----------------------------+
87+
* | encoded integers
88+
* v
89+
* +------------------------------------------------------------------+
90+
* | 3. DECODE |
91+
* | * Apply inverse formula: value[i] = encoded[i] * 10^-e * 10^f |
92+
* +------------------------------------+-----------------------------+
93+
* | decoded floats (with placeholders)
94+
* v
95+
* +------------------------------------------------------------------+
96+
* | 4. PATCH EXCEPTIONS |
97+
* | * Replace values at exceptionPositions[] with exceptions[] |
98+
* +------------------------------------+-----------------------------+
99+
* |
100+
* v
101101
* Output: Original float/double array (lossless!)
102102
*
103-
* ═══════════════════════════════════════════════════════════════════════════════════════
103+
* ==========================================================================================
104104
*/
105105

106106
/**
@@ -130,21 +130,21 @@ struct AlpExponentAndFactor {
130130
*
131131
* Serialization format (stored as raw binary struct):
132132
*
133-
* ┌──────────────────────────────────────────┐
134-
* │ AlpEncodedVectorInfo (23+ bytes) │
135-
* ├──────────────────────────────────────────┤
136-
* │ Offset │ Field │ Size │
137-
* ├─────────┼─────────────────────┼──────────┤
138-
* │ 0 │ exponent (uint8_t) │ 1 byte │
139-
* │ 1 │ factor (uint8_t) │ 1 byte │
140-
* │ 2 │ [padding] │ 6 bytes │
141-
* │ 8 │ frameOfReference │ 8 bytes │
142-
* │ 16 │ bitWidth (uint8_t) │ 1 byte │
143-
* │ 17 │ [padding] │ 7 bytes │
144-
* │ 24 │ bitPackedSize │ 8 bytes │
145-
* │ 32 │ numElements │ 2 bytes │
146-
* │ 34 │ numExceptions │ 2 bytes │
147-
* └──────────────────────────────────────────┘
133+
* +------------------------------------------+
134+
* | AlpEncodedVectorInfo (23+ bytes) |
135+
* +------------------------------------------+
136+
* | Offset | Field | Size |
137+
* +---------+---------------------+----------+
138+
* | 0 | exponent (uint8_t) | 1 byte |
139+
* | 1 | factor (uint8_t) | 1 byte |
140+
* | 2 | [padding] | 6 bytes |
141+
* | 8 | frameOfReference | 8 bytes |
142+
* | 16 | bitWidth (uint8_t) | 1 byte |
143+
* | 17 | [padding] | 7 bytes |
144+
* | 24 | bitPackedSize | 8 bytes |
145+
* | 32 | numElements | 2 bytes |
146+
* | 34 | numExceptions | 2 bytes |
147+
* +------------------------------------------+
148148
*/
149149
struct AlpEncodedVectorInfo {
150150
/// Exponent and factor used for compression.
@@ -174,23 +174,23 @@ struct AlpEncodedVectorInfo {
174174
/**
175175
* Complete serialization format for an ALP compressed vector:
176176
*
177-
* ┌────────────────────────────────────────────────────────────┐
178-
* │ AlpEncodedVector<T> Serialized Layout │
179-
* ├────────────────────────────────────────────────────────────┤
180-
* │ Section │ Size (bytes) │ Description │
181-
* ├───────────────────────┼──────────────────────┼──────────────┤
182-
* │ 1. VectorInfo │ sizeof(VectorInfo) │ Metadata │
183-
* │ (see above) │ (~36 with padding) │ │
184-
* ├───────────────────────┼──────────────────────┼──────────────┤
185-
* │ 2. Packed Values │ bitPackedSize │ Bitpacked │
186-
* │ (compressed data) │ (variable) │ integers │
187-
* ├───────────────────────┼──────────────────────┼──────────────┤
188-
* │ 3. Exception Pos │ numExceptions * 2 │ uint16_t[] │
189-
* │ (indices) │ (variable) │ positions │
190-
* ├───────────────────────┼──────────────────────┼──────────────┤
191-
* │ 4. Exception Values │ numExceptions * │ T[] (float/ │
192-
* │ (original floats) │ sizeof(T) │ double) │
193-
* └────────────────────────────────────────────────────────────┘
177+
* +------------------------------------------------------------+
178+
* | AlpEncodedVector<T> Serialized Layout |
179+
* +------------------------------------------------------------+
180+
* | Section | Size (bytes) | Description |
181+
* +-----------------------+----------------------+-------------+
182+
* | 1. VectorInfo | sizeof(VectorInfo) | Metadata |
183+
* | (see above) | (~36 with padding) | |
184+
* +-----------------------+----------------------+-------------+
185+
* | 2. Packed Values | bitPackedSize | Bitpacked |
186+
* | (compressed data) | (variable) | integers |
187+
* +-----------------------+----------------------+-------------+
188+
* | 3. Exception Pos | numExceptions * 2 | uint16_t[] |
189+
* | (indices) | (variable) | positions |
190+
* +-----------------------+----------------------+-------------+
191+
* | 4. Exception Values | numExceptions * | T[] (float/|
192+
* | (original floats) | sizeof(T) | double) |
193+
* +------------------------------------------------------------+
194194
*
195195
* Example for 1024 floats with 5 exceptions and bitWidth=8:
196196
* - VectorInfo: 36 bytes

0 commit comments

Comments
 (0)