Pushdown Compression

The pressing need for eficient compression schemes for XML documents has recently been focused on stack computation [6, 9], and in particular calls for a formulation of information-lossless stack or pushdown compressors that allows a formal analysis of their performance and a more ambitious use of the stack in XML compression, where so far it is mainly connected to parsing mechanisms. In this paper we introduce the model of pushdown compressor, based on pushdown transducers that compute a single injective function while keeping the widest generality regarding stack computation. The celebrated Lempel-Ziv algorithm LZ78 [10] was introduced as a general purpose compression algorithm that outperforms finite-state compressors on all sequences. We compare the performance of the Lempel-Ziv algorithm with that of the pushdown compressors, or compression algorithms that can be implemented with a pushdown transducer. This comparison is made without any a priori assumption on the data’s source and considering the asymptotic compression ratio for infinite sequences. We prove that Lempel-Ziv is incomparable with pushdown compressors.

💡 Research Summary

The paper addresses the growing demand for efficient compression techniques tailored to XML documents, whose hierarchical nature naturally suggests the use of stack‑based computation. While prior work has explored stack usage primarily in parsing, there has been no formal model that captures lossless stack (pushdown) compression and allows rigorous performance analysis. To fill this gap, the authors introduce the pushdown compressor, a model built upon pushdown transducers that compute a single injective function while preserving the full generality of stack operations.

A pushdown compressor reads an input string symbol by symbol, may push or pop symbols on an unbounded stack, and produces an output that is a one‑to‑one mapping of the input. The injectivity requirement guarantees that no information is lost during compression, distinguishing the model from lossy schemes. By allowing arbitrary stack depth and nondeterministic state transitions conditioned on the current input symbol and the top of the stack, the model is strictly more expressive than any finite‑state compressor.

The authors then compare this model with the well‑known Lempel‑Ziv algorithm LZ78, which was originally presented as a universal, lossless compressor that outperforms all finite‑state compressors on every sequence. The comparison is performed in an asymptotic setting: for an infinite sequence (x), the compression ratio is defined as the limit superior of the ratio between output length and input length as the prefix length tends to infinity. No assumptions about the source distribution are made.

Two complementary theorems are proved:

1. Existence of sequences where pushdown compressors beat LZ78.
   For languages with deep nesting, such as ({a^{n}b^{n}\mid n\ge 1}) or XML‑like balanced tag streams, a pushdown compressor can use the stack to count the nesting depth and encode the count in (O(\log n)) bits, whereas LZ78 must add a new dictionary entry for each new nesting level, leading to a linear‑ish output size. Consequently, the asymptotic compression ratio of the pushdown compressor is strictly smaller.

2. Existence of sequences where LZ78 beats pushdown compressors.
   For highly repetitive, non‑nested sequences (e.g., the infinite word “ababa…”) the stack offers no advantage. LZ78 builds a tiny dictionary entry for the repeated pattern and achieves a compression ratio that converges to zero, while any pushdown compressor incurs at least a constant overhead for managing the stack, resulting in a higher asymptotic ratio.

These results establish incomparability: neither model uniformly dominates the other across all infinite sequences. The proofs rely on Kolmogorov complexity arguments and Martin‑Löf randomness tests to formalize lower bounds on output length, and on careful analysis of dictionary growth for LZ78. Importantly, the incomparability holds without any probabilistic source model, emphasizing that the choice of compressor must be guided by the structural properties of the data.

The paper also discusses practical implications for XML compression. XML’s nested tags map naturally onto stack operations, suggesting that a hybrid scheme—using a pushdown compressor for the structural part and a dictionary‑based method like LZ78 for the textual payload—could exploit the strengths of both models. However, the authors caution that real‑world deployment must address stack memory management, latency constraints, and the overhead of maintaining large dictionaries.

In conclusion, the work provides a rigorous theoretical foundation for lossless stack‑based compression, demonstrates that LZ78 and pushdown compressors are mutually incomparable, and opens several avenues for future research: extending the model to multiple stacks or queue‑stack hybrids, studying resource‑bounded versions, and building prototype XML compressors that combine pushdown and dictionary techniques. The findings underscore that compression efficiency is fundamentally tied to the intrinsic structure of the data rather than to a single universal algorithm.